WEBVTT

00:00:00.240 --> 00:00:03.980
<v Michael Kennedy>Should your agent call an MCP server or just run a CLI?

00:00:04.540 --> 00:00:09.320
<v Michael Kennedy>Pamela Fox joins us to share how she thinks about connecting AIs and her data.

00:00:09.960 --> 00:00:13.100
<v Michael Kennedy>She's taught MCP workshops from PyCon to VR developers,

00:00:13.480 --> 00:00:15.940
<v Michael Kennedy>and every week an agent combs her Gmail,

00:00:16.880 --> 00:00:19.480
<v Michael Kennedy>Outlook, and X history to build her news roundup.

00:00:20.000 --> 00:00:21.220
<v Michael Kennedy>She uses MCP all day,

00:00:21.760 --> 00:00:23.760
<v Michael Kennedy>yet there are still times when you should skip it.

00:00:24.240 --> 00:00:27.140
<v Michael Kennedy>You'll hear how her colleague pitted the GitHub CLI

00:00:27.180 --> 00:00:30.000
<v Michael Kennedy>against the GitHub MCP server on a real bug.

00:00:30.400 --> 00:00:31.660
<v Michael Kennedy>The CLI used fewer tokens.

00:00:32.140 --> 00:00:33.180
<v Michael Kennedy>It succeeded more often.

00:00:33.940 --> 00:00:35.720
<v Michael Kennedy>So do we choose CLIs,

00:00:36.020 --> 00:00:36.580
<v Michael Kennedy>MCPs,

00:00:36.780 --> 00:00:37.740
<v Michael Kennedy>or even just APIs

00:00:38.000 --> 00:00:40.520
<v Michael Kennedy>for connecting our agents to our data and services?

00:00:41.200 --> 00:00:41.800
<v Michael Kennedy>Let's find out.

00:00:42.180 --> 00:00:44.440
<v Michael Kennedy>This is Talk Python To Me, episode 569,

00:00:44.540 --> 00:00:47.160
<v Michael Kennedy>recorded October 15th, 2026.

00:01:04.320 --> 00:01:09.000
<v Michael Kennedy>Welcome to Talk Python To Me, the number one Python podcast for developers and data scientists.

00:01:09.520 --> 00:01:10.920
<v Michael Kennedy>This is your host, Michael Kennedy.

00:01:11.120 --> 00:01:14.840
<v Michael Kennedy>I'm a PSF fellow who's been coding for over 25 years.

00:01:15.260 --> 00:01:16.540
<v Michael Kennedy>Let's connect on social media.

00:01:16.980 --> 00:01:19.180
<v Michael Kennedy>You'll find me and Talk Python on Mastodon, Bluesky,

00:01:19.560 --> 00:01:20.000
<v Michael Kennedy>and X.

00:01:20.250 --> 00:01:22.180
<v Michael Kennedy>The social links are all in your show notes.

00:01:22.600 --> 00:01:25.540
<v Michael Kennedy>You can find over 10 years of past episodes at

00:01:25.660 --> 00:01:26.460
<v Michael Kennedy>Talk Python.fm.

00:01:26.740 --> 00:01:29.840
<v Michael Kennedy>And if you want to be part of the show, you can join our recording live streams.

00:01:30.260 --> 00:01:34.120
<v Michael Kennedy>That's right. We live stream the raw uncut version of each episode on YouTube.

00:01:34.780 --> 00:01:35.160
<v Michael Kennedy>Just visit

00:01:35.300 --> 00:01:39.140
<v Michael Kennedy>Talk Python.fm slash YouTube to see the schedule of upcoming events.

00:01:39.330 --> 00:01:40.480
<v Michael Kennedy>Be sure to subscribe there and

00:01:40.490 --> 00:01:43.020
<v Michael Kennedy>press the bell so you'll get notified anytime we're recording.

00:01:43.620 --> 00:01:45.020
<v Michael Kennedy>Before we dive into the show,

00:01:45.200 --> 00:01:49.220
<v Michael Kennedy>I want to tell you about an upcoming members meetup here at Talk Python.

00:01:49.960 --> 00:01:50.420
<v Michael Kennedy>The title,

00:01:50.700 --> 00:01:54.140
<v Michael Kennedy>SQLite Might Actually Be All You Need.

00:01:54.840 --> 00:01:56.780
<v Michael Kennedy>This talk dives into running SQLite in production

00:01:56.900 --> 00:01:57.980
<v Michael Kennedy>and why, surprisingly,

00:01:58.340 --> 00:02:02.260
<v Michael Kennedy>it might be one of the best production databases you can choose.

00:02:03.120 --> 00:02:04.960
<v Michael Kennedy>Here's something I haven't really talked about yet.

00:02:05.460 --> 00:02:07.500
<v Michael Kennedy>Talk Python doesn't run on MongoDB anymore.

00:02:08.039 --> 00:02:09.080
<v Michael Kennedy>After years on Mongo,

00:02:09.300 --> 00:02:12.260
<v Michael Kennedy>I've moved almost all of our sites to SQLite,

00:02:12.580 --> 00:02:14.220
<v Michael Kennedy>the database that you might file under.

00:02:14.680 --> 00:02:16.040
<v Michael Kennedy>Fine for prototypes and phone apps.

00:02:16.680 --> 00:02:19.760
<v Michael Kennedy>I knew an in-process database would have some speed advantages,

00:02:20.100 --> 00:02:25.200
<v Michael Kennedy>but what I didn't expect is that it's 40 to 300 times faster on our data.

00:02:26.020 --> 00:02:26.240
<v Michael Kennedy>Now,

00:02:26.400 --> 00:02:28.920
<v Michael Kennedy>looking up a guest takes 1.5 microseconds.

00:02:29.180 --> 00:02:31.560
<v Michael Kennedy>That's 1 1,000th of a millisecond,

00:02:31.760 --> 00:02:35.320
<v Michael Kennedy>or over 600,000 calls a second from plain Python.

00:02:36.120 --> 00:02:36.640
<v Michael Kennedy>And a write?

00:02:37.220 --> 00:02:39.540
<v Michael Kennedy>Well, that's 1 1 100th of a millisecond.

00:02:40.260 --> 00:02:42.620
<v Michael Kennedy>But then there's that question that scares people off.

00:02:42.700 --> 00:02:43.920
<v Michael Kennedy>What happens when the server dies?

00:02:44.580 --> 00:02:46.240
<v Michael Kennedy>This is where the surprises live.

00:02:47.000 --> 00:02:49.740
<v Michael Kennedy>So I hope you join me for the members tech meetup.

00:02:49.940 --> 00:02:51.740
<v Michael Kennedy>It's on December 2nd at 10 a.m.

00:02:52.000 --> 00:02:53.020
<v Michael Kennedy>U.S. Pacific time.

00:02:53.740 --> 00:02:55.700
<v Michael Kennedy>Find all the details at Talk Python.

00:02:55.850 --> 00:02:58.160
<v Michael Kennedy>Just click membership and go to the tech talks.

00:02:58.820 --> 00:02:59.920
<v Michael Kennedy>Add it to your calendar right there.

00:03:00.050 --> 00:03:03.260
<v Michael Kennedy>I hope to see you at the meetup and share some SQLite surprises with you.

00:03:04.020 --> 00:03:04.380
<v Michael Kennedy>Pamela,

00:03:04.540 --> 00:03:06.080
<v Michael Kennedy>welcome to Talk Python To Me.

00:03:06.280 --> 00:03:08.580
<v Michael Kennedy>It is long past time I had you on the show.

00:03:08.790 --> 00:03:09.440
<v Michael Kennedy>Really glad you're here.

00:03:10.000 --> 00:03:10.360
<v Pamela Fox>Awesome.

00:03:10.860 --> 00:03:11.560
<v Pamela Fox>It's so fun to be here.

00:03:11.800 --> 00:03:17.220
<v Michael Kennedy>We're going to talk about some of the craziness of 2026 and choices.

00:03:18.010 --> 00:03:20.980
<v Michael Kennedy>Should we MCP to MCP or not?

00:03:21.140 --> 00:03:23.440
<v Michael Kennedy>Was that the Shakespeare question? I think that was it.

00:03:23.500 --> 00:03:23.940
<v Pamela Fox>Yeah, exactly.

00:03:24.170 --> 00:03:25.960
<v Michael Kennedy>Or the CLI or APIs,

00:03:26.310 --> 00:03:32.320
<v Michael Kennedy>or basically I've got some data and services and I want to connect them to my code, to my agents.

00:03:32.580 --> 00:03:33.520
<v Michael Kennedy>How do we do that, right?

00:03:33.580 --> 00:03:39.780
<v Michael Kennedy>I know you've written a whole lot about this and people should absolutely check out your blog, but especially conference talks.

00:03:40.320 --> 00:03:43.160
<v Michael Kennedy>You're an awesome technologist and I really loved all of your presentations,

00:03:43.560 --> 00:03:45.700
<v Michael Kennedy>you know, including the one just at PyBay where we were.

00:03:45.820 --> 00:03:47.580
<v Michael Kennedy>How was PyBay? What do you think?

00:03:48.780 --> 00:03:50.780
<v Pamela Fox>I thought PyBay was really fun.

00:03:51.120 --> 00:03:53.220
<v Pamela Fox>I love the Python conferences,

00:03:53.780 --> 00:03:57.880
<v Pamela Fox>getting to meet the local community and you had a great talk at it as well.

00:03:58.140 --> 00:03:58.480
<v Michael Kennedy>Oh, yeah.

00:03:58.700 --> 00:04:00.840
<v Michael Kennedy>Thanks. That was really fun about Talk Python in production.

00:04:00.880 --> 00:04:02.000
<v Michael Kennedy>that will come out eventually.

00:04:02.280 --> 00:04:04.480
<v Michael Kennedy>I'm sure the videos of those will come out. They were recorded,

00:04:04.700 --> 00:04:06.100
<v Michael Kennedy>so they wouldn't record them for no reason.

00:04:07.280 --> 00:04:07.680
<v Pamela Fox>Yes.

00:04:08.360 --> 00:04:10.120
<v Michael Kennedy>So really, really cool. Now, before we dive in,

00:04:10.460 --> 00:04:11.060
<v Michael Kennedy>as per usual,

00:04:11.260 --> 00:04:12.920
<v Michael Kennedy>just give people a quick background on you.

00:04:13.200 --> 00:04:13.300
<v Pamela Fox>Yeah,

00:04:13.760 --> 00:04:13.900
<v Pamela Fox>sure.

00:04:14.100 --> 00:04:15.940
<v Pamela Fox>So I am Pamela Fox. I am

00:04:16.260 --> 00:04:19.400
<v Pamela Fox>currently at Microsoft and or GitHub,

00:04:20.120 --> 00:04:24.000
<v Pamela Fox>whichever company you prefer more.

00:04:24.420 --> 00:04:25.580
<v Pamela Fox>And I am a principal

00:04:25.840 --> 00:04:26.420
<v Pamela Fox>cloud advocate.

00:04:26.900 --> 00:04:28.380
<v Pamela Fox>So that means that generally,

00:04:28.540 --> 00:04:30.279
<v Pamela Fox>like I'm using all of our developer tools

00:04:30.300 --> 00:04:32.620
<v Pamela Fox>and showing developers how to use them.

00:04:33.520 --> 00:04:35.180
<v Pamela Fox>I tend to focus on Python.

00:04:35.720 --> 00:04:39.860
<v Pamela Fox>We had like an official Python advocacy team for four years

00:04:40.280 --> 00:04:41.860
<v Pamela Fox>and we don't anymore,

00:04:42.790 --> 00:04:43.380
<v Pamela Fox>which is sad.

00:04:43.380 --> 00:04:46.540
<v Pamela Fox>And that's kind of part of the change due to coding agents

00:04:46.880 --> 00:04:49.760
<v Pamela Fox>where like languages are not as much of a thing anymore

00:04:50.360 --> 00:04:52.880
<v Pamela Fox>because the coding agents can just code and whatever

00:04:53.500 --> 00:04:54.920
<v Pamela Fox>and all that matters is the output.

00:04:55.640 --> 00:04:56.060
<v Pamela Fox>So now,

00:04:56.500 --> 00:04:56.760
<v Pamela Fox>you know,

00:04:57.120 --> 00:04:58.739
<v Pamela Fox>I guess there's just this big question

00:04:58.760 --> 00:05:01.880
<v Pamela Fox>as to how much does your language even matter anymore.

00:05:02.010 --> 00:05:04.480
<v Pamela Fox>So I'm glad on that note with PyBay,

00:05:04.880 --> 00:05:06.320
<v Pamela Fox>I currently think it's great

00:05:06.520 --> 00:05:08.520
<v Pamela Fox>that we still have these Python-specific conferences.

00:05:09.060 --> 00:05:12.000
<v Pamela Fox>It will be interesting to see how long language

00:05:12.600 --> 00:05:14.920
<v Pamela Fox>is the theme that holds together a community.

00:05:15.240 --> 00:05:16.420
<v Michael Kennedy>What an interesting idea.

00:05:16.520 --> 00:05:18.740
<v Michael Kennedy>I've been thinking about this a lot too, and I fright.

00:05:19.320 --> 00:05:21.400
<v Michael Kennedy>I was, before we go down this AI thing,

00:05:22.440 --> 00:05:23.580
<v Michael Kennedy>when you were talking about

00:05:24.060 --> 00:05:26.400
<v Michael Kennedy>not specifically having a Python team,

00:05:26.900 --> 00:05:27.580
<v Michael Kennedy>one of the thoughts I had,

00:05:27.700 --> 00:05:29.080
<v Michael Kennedy>well, like Python is just so popular,

00:05:29.240 --> 00:05:30.080
<v Michael Kennedy>just it's expected.

00:05:30.540 --> 00:05:32.620
<v Michael Kennedy>It's not this niche thing where we brought in a few people

00:05:32.780 --> 00:05:34.980
<v Michael Kennedy>that know this weird no curly braces language.

00:05:35.580 --> 00:05:36.020
<v Michael Kennedy>You know what I mean?

00:05:36.160 --> 00:05:38.940
<v Michael Kennedy>But like just the advocates know about Python.

00:05:39.420 --> 00:05:41.080
<v Michael Kennedy>And so maybe you don't need a specific Python team.

00:05:41.360 --> 00:05:44.880
<v Pamela Fox>Yeah, I guess two things happen is one is that more and more

00:05:45.200 --> 00:05:46.980
<v Pamela Fox>the advocates from the other language teams

00:05:47.440 --> 00:05:50.460
<v Pamela Fox>found themselves using Python because you almost have to now

00:05:50.600 --> 00:05:54.400
<v Pamela Fox>because that's like it became the default language for new SDKs, right?

00:05:54.480 --> 00:05:56.400
<v Pamela Fox>When I first joined Marksoft, it was the opposite, right?

00:05:56.500 --> 00:05:58.220
<v Pamela Fox>where we had to like kind of beg for a Python SDK.

00:05:59.140 --> 00:06:02.740
<v Pamela Fox>And then people had to beg for like SDKs in C#,

00:06:02.950 --> 00:06:04.140
<v Pamela Fox>which is a crazy thing for Microsoft

00:06:04.460 --> 00:06:05.720
<v Pamela Fox>because Microsoft is like a huge,

00:06:05.940 --> 00:06:06.800
<v Pamela Fox>you know, C# user.

00:06:07.260 --> 00:06:08.780
<v Michael Kennedy>Like they may have created the language.

00:06:09.700 --> 00:06:09.900
<v Michael Kennedy>Right,

00:06:10.200 --> 00:06:10.820
<v Pamela Fox>that thing.

00:06:10.980 --> 00:06:11.340
<v Pamela Fox>Yes, yes,

00:06:11.580 --> 00:06:11.860
<v Pamela Fox>that's true.

00:06:12.840 --> 00:06:14.360
<v Pamela Fox>So yeah, so I think part of it

00:06:14.460 --> 00:06:16.180
<v Pamela Fox>is that just basically all advocates,

00:06:16.700 --> 00:06:16.900
<v Pamela Fox>you know,

00:06:17.080 --> 00:06:18.020
<v Pamela Fox>they're, you know, we're all there.

00:06:18.120 --> 00:06:19.500
<v Pamela Fox>Everyone's using Python now,

00:06:19.900 --> 00:06:20.860
<v Pamela Fox>all using JavaScript now.

00:06:21.040 --> 00:06:21.720
<v Pamela Fox>Everyone's using every,

00:06:21.880 --> 00:06:21.980
<v Pamela Fox>you know,

00:06:22.040 --> 00:06:24.020
<v Pamela Fox>because it's just so easy to add on,

00:06:24.260 --> 00:06:25.240
<v Pamela Fox>you know, a React front end.

00:06:25.440 --> 00:06:27.180
<v Pamela Fox>I saw a talk from my dad the other day.

00:06:27.340 --> 00:06:28.380
<v Pamela Fox>He's a computer scientist.

00:06:29.300 --> 00:06:31.940
<v Pamela Fox>And on his slide, he had like React front end.

00:06:31.950 --> 00:06:32.420
<v Pamela Fox>I was like, what?

00:06:32.960 --> 00:06:33.100
<v Pamela Fox>Dad,

00:06:33.560 --> 00:06:34.220
<v Pamela Fox>you're doing React?

00:06:34.480 --> 00:06:35.300
<v Pamela Fox>Like what has happened,

00:06:35.560 --> 00:06:35.620
<v Pamela Fox>right?

00:06:35.720 --> 00:06:38.560
<v Pamela Fox>That's only due to coding agents.

00:06:38.670 --> 00:06:39.360
<v Pamela Fox>I guarantee it.

00:06:39.430 --> 00:06:44.420
<v Pamela Fox>Because there's no way that my dad would be just like suddenly like, well, I'm like seven

00:06:44.480 --> 00:06:44.800
<v Pamela Fox>years old.

00:06:44.850 --> 00:06:46.500
<v Pamela Fox>I think it's time that I learned React,

00:06:46.860 --> 00:06:46.960
<v Pamela Fox>right?

00:06:47.240 --> 00:06:47.580
<v Michael Kennedy>Exactly.

00:06:47.700 --> 00:06:47.940
<v Michael Kennedy>It's like,

00:06:48.340 --> 00:06:49.860
<v Michael Kennedy>Pamela, what are you using Node for?

00:06:49.980 --> 00:06:50.980
<v Michael Kennedy>We all use Bun now.

00:06:51.140 --> 00:06:51.460
<v Michael Kennedy>Come on.

00:06:52.500 --> 00:06:52.740
<v Michael Kennedy>What?

00:06:53.000 --> 00:06:53.640
<v Michael Kennedy>Who are you?

00:06:54.180 --> 00:06:54.340
<v Pamela Fox>Yeah.

00:06:54.640 --> 00:06:57.000
<v Pamela Fox>I need to call him up so we can talk JavaScript now.

00:06:57.180 --> 00:06:57.360
<v Michael Kennedy>Funny.

00:06:57.540 --> 00:06:57.640
<v Michael Kennedy>Yeah,

00:06:57.820 --> 00:06:58.140
<v Michael Kennedy>you're right.

00:06:58.360 --> 00:07:03.040
<v Michael Kennedy>And I honestly don't even know if I want to talk about it because it makes me a little bit sad.

00:07:03.140 --> 00:07:07.060
<v Michael Kennedy>But I do think that there is also something to the second half of the reason that you gave.

00:07:07.480 --> 00:07:10.480
<v Michael Kennedy>And that's we don't necessarily have to read the code.

00:07:10.880 --> 00:07:11.440
<v Michael Kennedy>Should we read the code?

00:07:11.480 --> 00:07:12.260
<v Michael Kennedy>I don't know.

00:07:12.380 --> 00:07:13.400
<v Michael Kennedy>I'm very conflicted.

00:07:13.400 --> 00:07:17.880
<v Michael Kennedy>I actually was thinking today about building a little GUI application,

00:07:18.460 --> 00:07:19.340
<v Michael Kennedy>plug into an LLM,

00:07:19.720 --> 00:07:20.640
<v Michael Kennedy>maybe an MCP, who knows,

00:07:20.920 --> 00:07:26.920
<v Michael Kennedy>that would basically go to a code base and go from this git commit to that one i need i need a tour

00:07:27.060 --> 00:07:32.600
<v Michael Kennedy>of what has happened will you explain this to me because git history is not there's a thousand

00:07:32.840 --> 00:07:35.900
<v Michael Kennedy>change files what is significant what do i you know what i mean something like that would be cool

00:07:36.260 --> 00:07:40.560
<v Pamela Fox>yeah yeah we care more about the high level changes and also just how they're being verified

00:07:41.020 --> 00:07:47.999
<v Pamela Fox>so i think like in terms of python you know what i am most interested in is you know the verifiability

00:07:48.020 --> 00:07:49.880
<v Pamela Fox>We do have great testing frameworks in Python.

00:07:50.120 --> 00:07:51.200
<v Pamela Fox>So I think we can continue,

00:07:51.780 --> 00:07:52.080
<v Pamela Fox>you know,

00:07:52.580 --> 00:07:54.260
<v Pamela Fox>talking about those testing frameworks,

00:07:54.600 --> 00:07:55.020
<v Pamela Fox>like how you

00:07:55.180 --> 00:08:00.080
<v Pamela Fox>could encourage your agents to use PlayWrite and use pytest and use hypothesis and schema

00:08:00.260 --> 00:08:01.860
<v Pamela Fox>thesis and all of that stuff.

00:08:02.100 --> 00:08:02.380
<v Pamela Fox>And,

00:08:02.640 --> 00:08:06.560
<v Pamela Fox>you know, at least guide it to using like an async FastAPI,

00:08:07.140 --> 00:08:08.140
<v Pamela Fox>you know, versus a

00:08:08.340 --> 00:08:08.840
<v Pamela Fox>sync framework.

00:08:10.420 --> 00:08:12.340
<v Pamela Fox>But the specifics of Python,

00:08:13.080 --> 00:08:13.340
<v Pamela Fox>you know,

00:08:13.580 --> 00:08:15.060
<v Pamela Fox>not as useful.

00:08:15.300 --> 00:08:19.100
<v Pamela Fox>I was reading the Python 315 release notes and I was like, oh, all of this is great.

00:08:19.300 --> 00:08:19.640
<v Pamela Fox>I'm glad.

00:08:19.840 --> 00:08:20.140
<v Pamela Fox>I know.

00:08:20.140 --> 00:08:21.240
<v Pamela Fox>I'm glad they added this.

00:08:21.660 --> 00:08:26.320
<v Pamela Fox>But none of this matters until the LLMs know it.

00:08:26.920 --> 00:08:27.200
<v Michael Kennedy>I see.

00:08:27.620 --> 00:08:32.659
<v Pamela Fox>Until they've been pointed at the release notes and until I'm using 315 in production.

00:08:32.880 --> 00:08:34.219
<v Michael Kennedy>We're going to see what the world brings.

00:08:34.440 --> 00:08:36.539
<v Michael Kennedy>But it is trending this way and it is weird.

00:08:36.919 --> 00:08:40.700
<v Michael Kennedy>I don't think that means that people without programming skills are not needed or don't

00:08:40.700 --> 00:08:41.140
<v Michael Kennedy>have value.

00:08:41.400 --> 00:08:42.320
<v Michael Kennedy>I think there's tons.

00:08:42.380 --> 00:08:45.200
<v Michael Kennedy>It's just not in the structure level.

00:08:45.420 --> 00:08:45.700
<v Michael Kennedy>It's like, okay,

00:08:45.960 --> 00:08:47.200
<v Michael Kennedy>we need this kind of design pattern.

00:08:47.260 --> 00:08:48.560
<v Michael Kennedy>We need this kind of storage.

00:08:49.160 --> 00:08:50.840
<v Michael Kennedy>What are we going to do about, you know,

00:08:51.020 --> 00:08:54.240
<v Michael Kennedy>durability if this data center goes down and we got our data there?

00:08:54.560 --> 00:08:55.340
<v Michael Kennedy>That kind of questions,

00:08:55.520 --> 00:08:55.620
<v Michael Kennedy>yeah?

00:08:55.920 --> 00:08:57.440
<v Pamela Fox>Yeah, and that's all things we've learned.

00:08:57.600 --> 00:09:01.220
<v Pamela Fox>You know, syntax was just one way that we expressed our programming knowledge.

00:09:01.860 --> 00:09:02.200
<v Pamela Fox>Now,

00:09:02.520 --> 00:09:06.260
<v Pamela Fox>you know, we're really going to express the architecture side of it because that's where,

00:09:06.640 --> 00:09:06.760
<v Pamela Fox>you know,

00:09:06.800 --> 00:09:09.660
<v Pamela Fox>we still need to think a lot about how we're going to architecture things,

00:09:09.820 --> 00:09:10.780
<v Pamela Fox>what the protocols are

00:09:11.200 --> 00:09:13.460
<v Pamela Fox>that we're going to use in order to interface across systems,

00:09:13.940 --> 00:09:15.020
<v Pamela Fox>which that's like a good lead

00:09:15.020 --> 00:09:16.320
<v Pamela Fox>in for today's topic,

00:09:16.700 --> 00:09:16.820
<v Pamela Fox>because,

00:09:17.040 --> 00:09:20.000
<v Pamela Fox>you know, that I think the protocols are increasingly the important

00:09:20.200 --> 00:09:20.340
<v Pamela Fox>part.

00:09:20.520 --> 00:09:20.660
<v Michael Kennedy>Absolutely.

00:09:21.220 --> 00:09:23.380
<v Michael Kennedy>All right. Well, let's go ahead and dive into it. I think,

00:09:23.820 --> 00:09:24.580
<v Michael Kennedy>as I said, I kind of

00:09:24.640 --> 00:09:29.640
<v Michael Kennedy>want to structure this around MCP because that's maybe the most obvious choice. Somebody says,

00:09:29.760 --> 00:09:32.120
<v Michael Kennedy>I want to connect this thing to my AI. Well,

00:09:32.320 --> 00:09:33.740
<v Michael Kennedy>MCP were obviously right.

00:09:34.020 --> 00:09:34.140
<v Michael Kennedy>However,

00:09:34.820 --> 00:09:35.339
<v Michael Kennedy>there's a lot of

00:09:35.320 --> 00:09:35.480
<v Michael Kennedy>options,

00:09:35.940 --> 00:09:36.420
<v Michael Kennedy>as we'll see.

00:09:36.620 --> 00:09:41.020
<v Michael Kennedy>And there's times you do and don't use MCPs. But let's start with that

00:09:41.240 --> 00:09:41.400
<v Michael Kennedy>story.

00:09:41.660 --> 00:09:47.260
<v Michael Kennedy>So I had a Den Bilomarski on the show a while ago to talk about MCP protocol.

00:09:47.580 --> 00:09:48.240
<v Michael Kennedy>But honestly,

00:09:48.620 --> 00:09:51.300
<v Michael Kennedy>it's like a year ago and in AI world,

00:09:51.660 --> 00:09:53.460
<v Michael Kennedy>you know, how crazy, how long ago was that?

00:09:53.740 --> 00:09:54.240
<v Michael Kennedy>So maybe we

00:09:54.280 --> 00:09:55.840
<v Michael Kennedy>should just start from the beginning.

00:09:56.040 --> 00:09:57.520
<v Michael Kennedy>You know, what is an MCP?

00:09:58.440 --> 00:09:59.680
<v Michael Kennedy>What is MCP? And what is an MCP

00:09:59.920 --> 00:10:00.000
<v Michael Kennedy>server?

00:10:00.260 --> 00:10:01.240
<v Michael Kennedy>Like, why do we care about this stuff?

00:10:01.520 --> 00:10:01.580
<v Pamela Fox>Right.

00:10:01.880 --> 00:10:03.520
<v Pamela Fox>So MCP is,

00:10:03.700 --> 00:10:04.739
<v Pamela Fox>you know, it's the standard

00:10:04.740 --> 00:10:08.320
<v Pamela Fox>for helping AI applications to discover context,

00:10:09.400 --> 00:10:12.120
<v Pamela Fox>or here it says to connect to external systems.

00:10:13.420 --> 00:10:15.080
<v Pamela Fox>We basically wanted a standard way

00:10:15.200 --> 00:10:17.360
<v Pamela Fox>where we could extend our agents

00:10:17.900 --> 00:10:20.120
<v Pamela Fox>with access to additional tools,

00:10:20.340 --> 00:10:20.420
<v Pamela Fox>right?

00:10:20.480 --> 00:10:22.580
<v Pamela Fox>So it's extensibility story for agents,

00:10:22.920 --> 00:10:23.620
<v Pamela Fox>and it's a standard.

00:10:23.680 --> 00:10:24.820
<v Pamela Fox>And it's like, the cool thing is,

00:10:24.900 --> 00:10:26.020
<v Pamela Fox>it's like the first standard

00:10:26.240 --> 00:10:28.840
<v Pamela Fox>that everybody jumped on board with, right?

00:10:28.880 --> 00:10:31.120
<v Pamela Fox>It is so hard to get all these companies

00:10:31.440 --> 00:10:32.800
<v Pamela Fox>to agree right now,

00:10:33.060 --> 00:10:34.980
<v Pamela Fox>And they're like fighting on Twitter.

00:10:35.140 --> 00:10:36.400
<v Pamela Fox>It's like, come on, can't we all get along?

00:10:37.100 --> 00:10:38.080
<v Pamela Fox>And finally with MCP,

00:10:38.360 --> 00:10:39.080
<v Pamela Fox>you know, we can all get along.

00:10:39.320 --> 00:10:39.800
<v Pamela Fox>So this is,

00:10:40.040 --> 00:10:42.100
<v Pamela Fox>you know, originally came from Anthropic with Claude.

00:10:43.080 --> 00:10:45.320
<v Pamela Fox>But now there's great support in ChatGVT.

00:10:45.720 --> 00:10:48.400
<v Pamela Fox>You know, at Microsoft, we've got GitHub Copilot and VS Code.

00:10:49.580 --> 00:10:52.920
<v Pamela Fox>But there's also like consumer facing tools that have support for it.

00:10:53.180 --> 00:10:55.300
<v Pamela Fox>My business friends talk to me about MCP.

00:10:56.720 --> 00:10:57.120
<v Pamela Fox>Really?

00:10:57.280 --> 00:10:57.460
<v Michael Kennedy>Okay.

00:10:57.880 --> 00:10:58.040
<v Michael Kennedy>Yeah,

00:10:58.240 --> 00:10:58.360
<v Michael Kennedy>yeah.

00:10:59.140 --> 00:11:01.720
<v Pamela Fox>Their messaging made me like, oh, I just pitched this,

00:11:01.960 --> 00:11:02.180
<v Pamela Fox>you know,

00:11:02.280 --> 00:11:04.500
<v Pamela Fox>like extra using MCP.

00:11:04.920 --> 00:11:06.280
<v Pamela Fox>So yeah,

00:11:06.440 --> 00:11:07.000
<v Pamela Fox>because it's,

00:11:07.770 --> 00:11:07.960
<v Pamela Fox>you know,

00:11:08.140 --> 00:11:10.300
<v Pamela Fox>it's a standard that works across agents

00:11:10.560 --> 00:11:13.940
<v Pamela Fox>and agents are both developer facing and user agent.

00:11:14.160 --> 00:11:15.300
<v Pamela Fox>There's, you know, in the last few weeks,

00:11:15.480 --> 00:11:17.220
<v Pamela Fox>we've had GrokBot and Muse

00:11:18.810 --> 00:11:21.300
<v Pamela Fox>and there's so many new ones that pop out every week.

00:11:22.320 --> 00:11:26.320
<v Pamela Fox>And a lot of those are connecting behind the scenes using MCP.

00:11:26.350 --> 00:11:27.580
<v Pamela Fox>They don't always call it MCP.

00:11:27.800 --> 00:11:28.780
<v Pamela Fox>Chachapi calls it plugins,

00:11:29.190 --> 00:11:29.340
<v Pamela Fox>right?

00:11:30.220 --> 00:11:31.560
<v Pamela Fox>Like on their opening iDev day,

00:11:31.700 --> 00:11:33.980
<v Pamela Fox>They talked about ChattoBee plugins behind the scenes.

00:11:34.370 --> 00:11:37.820
<v Pamela Fox>Those are very much based off the MCP standard,

00:11:38.180 --> 00:11:38.280
<v Pamela Fox>right?

00:11:38.600 --> 00:11:38.780
<v Pamela Fox>So,

00:11:39.230 --> 00:11:46.620
<v Pamela Fox>you know, you're not always going to see MCP prominently displayed in the UI for these user-facing agents.

00:11:47.390 --> 00:11:53.820
<v Michael Kennedy>Yeah, I think if I go to Claude and I try to connect to it, it tells me what you need to do is add a connection,

00:11:54.340 --> 00:11:55.180
<v Michael Kennedy>not an MCP server,

00:11:55.360 --> 00:11:55.420
<v Michael Kennedy>right?

00:11:55.460 --> 00:11:57.260
<v Michael Kennedy>I'm like, this is where the MCP goes?

00:11:57.540 --> 00:11:57.940
<v Michael Kennedy>You know what I mean?

00:11:58.280 --> 00:11:58.480
<v Pamela Fox>Yeah,

00:11:58.680 --> 00:11:58.740
<v Pamela Fox>yeah.

00:11:58.780 --> 00:12:00.860
<v Pamela Fox>So I think they call it like connectors or connections.

00:12:01.700 --> 00:12:03.640
<v Pamela Fox>I'd be happy if they all call them MCP,

00:12:03.800 --> 00:12:06.460
<v Pamela Fox>but a lot of times they do wrap some additional things on top.

00:12:06.520 --> 00:12:07.400
<v Pamela Fox>Like ChatGPT,

00:12:08.300 --> 00:12:14.260
<v Pamela Fox>they wrap plugins in some additional metadata that helps them plug into the UI certain ways.

00:12:14.580 --> 00:12:17.980
<v Pamela Fox>So I think that's why they're calling it,

00:12:18.600 --> 00:12:21.220
<v Pamela Fox>they're coming up with their own name for it,

00:12:21.420 --> 00:12:23.160
<v Pamela Fox>just so they can add additional things on it.

00:12:23.420 --> 00:12:24.760
<v Michael Kennedy>Yeah, that makes a lot of sense.

00:12:25.140 --> 00:12:27.240
<v Michael Kennedy>And I guess it's like an API,

00:12:27.860 --> 00:12:27.960
<v Michael Kennedy>right?

00:12:28.300 --> 00:12:28.500
<v Pamela Fox>Right.

00:12:28.640 --> 00:12:29.900
<v Pamela Fox>Okay. So here's how it actually,

00:12:30.080 --> 00:12:32.200
<v Pamela Fox>like, I'll just tell you the actual, I mean, this is a developer,

00:12:32.560 --> 00:12:37.080
<v Pamela Fox>right? So what actually happens is first there's a handshake, right?

00:12:37.240 --> 00:12:43.720
<v Pamela Fox>So you register an MCP server and a lot of times your server is what's called an HTTP server. So there'll be actually like a URL for it.

00:12:44.460 --> 00:12:47.920
<v Pamela Fox>So it's like, you know, hpcoolserver.com slash MCP.

00:12:48.160 --> 00:12:51.880
<v Pamela Fox>And then the agent will connect to that server.

00:12:52.190 --> 00:12:54.100
<v Pamela Fox>So it'll send a handshake to that server.

00:12:54.270 --> 00:12:58.480
<v Pamela Fox>And during that handshake, it's going to ask the server what tools it has.

00:12:58.560 --> 00:13:02.540
<v Pamela Fox>because tools are like the there's other things that servers can expose but tools are like the

00:13:02.620 --> 00:13:09.420
<v Pamela Fox>really really important thing um so it'll say uh tools slash list so it sends this is a json rpc

00:13:09.420 --> 00:13:15.180
<v Pamela Fox>so it sends this json rpc uh you know hp requests to the server say hey what tools do you have

00:13:15.420 --> 00:13:20.360
<v Pamela Fox>and any other sort of discoverability it wants to do at that point mcp servers will support skills

00:13:20.620 --> 00:13:25.339
<v Pamela Fox>soon as well that's exciting um so it could discover those too so it discovers the tools

00:13:25.640 --> 00:13:26.700
<v Pamela Fox>the tools will come back.

00:13:26.900 --> 00:13:27.800
<v Pamela Fox>And once it has the tools,

00:13:27.980 --> 00:13:30.520
<v Pamela Fox>now the agent can populate its tools list

00:13:30.880 --> 00:13:32.020
<v Pamela Fox>with all the list of tools.

00:13:32.680 --> 00:13:37.060
<v Pamela Fox>And then it can decide to send those tool names to the LLM every time

00:13:37.160 --> 00:13:38.820
<v Pamela Fox>it's processing a user request.

00:13:39.340 --> 00:13:40.520
<v Pamela Fox>And if the LLM,

00:13:40.740 --> 00:13:42.000
<v Pamela Fox>the user request and says, oh,

00:13:42.240 --> 00:13:42.760
<v Pamela Fox>I've got a great

00:13:42.980 --> 00:13:44.180
<v Pamela Fox>tool for that, like book hotel,

00:13:44.720 --> 00:13:46.800
<v Pamela Fox>right? Maybe that's a tool and they want to book hotels.

00:13:47.480 --> 00:13:52.560
<v Pamela Fox>Then it's going to send a request to the MSP server that says it's like tools slash calls.

00:13:53.020 --> 00:13:57.280
<v Pamela Fox>And then it sends in arguments and says, OK, hotel in city, San Francisco,

00:13:57.640 --> 00:13:58.040
<v Pamela Fox>start date,

00:13:58.770 --> 00:14:01.000
<v Pamela Fox>October 15th, end date, October 17th.

00:14:01.190 --> 00:14:04.380
<v Pamela Fox>And so that sends, you know, basically like a remote procedure call like to the,

00:14:05.140 --> 00:14:09.840
<v Pamela Fox>you know, to that MSP server and the server processes it and responds back with,

00:14:10.000 --> 00:14:10.940
<v Pamela Fox>you know, JSON or,

00:14:11.240 --> 00:14:12.440
<v Pamela Fox>you know, however it wants to respond.

00:14:12.820 --> 00:14:16.020
<v Michael Kennedy>Yeah, it's a little bit like an HTTP endpoint,

00:14:16.380 --> 00:14:24.600
<v Michael Kennedy>but it's a little bit where the tools are, the individual functions that you might or the endpoints you might call with an API client.

00:14:25.020 --> 00:14:26.840
<v Michael Kennedy>But there's kind of a negotiation as well.

00:14:26.890 --> 00:14:28.060
<v Michael Kennedy>Like, do you support streaming?

00:14:29.180 --> 00:14:30.980
<v Michael Kennedy>What version of MCP are you?

00:14:31.260 --> 00:14:35.400
<v Michael Kennedy>So it's a little bit more high level than the HTTP APIs,

00:14:35.670 --> 00:14:35.800
<v Michael Kennedy>right?

00:14:35.960 --> 00:14:40.440
<v Michael Kennedy>Like maybe GraphQL or one of those types of something built on top of it, right?

00:14:40.500 --> 00:14:42.640
<v Michael Kennedy>But it's still very approachable for Python people.

00:14:42.960 --> 00:14:47.280
<v Pamela Fox>It's more opinionated because it's saying like, you know, it's saying this is specifically for agents.

00:14:47.880 --> 00:14:49.440
<v Pamela Fox>Agents want to know about tools.

00:14:49.660 --> 00:14:50.500
<v Pamela Fox>That's what they need.

00:14:50.900 --> 00:14:52.960
<v Pamela Fox>And they also want to know about, you know, in the future about skills.

00:14:53.260 --> 00:14:53.360
<v Pamela Fox>Right.

00:14:54.459 --> 00:14:54.860
<v Pamela Fox>But,

00:14:55.620 --> 00:15:00.080
<v Pamela Fox>you know, it's using mechanisms that we understand behind the scenes like JSON RPC.

00:15:00.600 --> 00:15:00.740
<v Michael Kennedy>Nice.

00:15:01.080 --> 00:15:04.080
<v Michael Kennedy>Yeah, I just last year, I can't remember exactly when it was.

00:15:04.080 --> 00:15:06.420
<v Michael Kennedy>I built a MCP server for the podcast.

00:15:07.060 --> 00:15:08.380
<v Michael Kennedy>So you can ask it like,

00:15:08.770 --> 00:15:10.900
<v Michael Kennedy>what did Talk Python cover on this?

00:15:11.040 --> 00:15:15.660
<v Michael Kennedy>And instead of having to go and do a full search of all the pages of the website,

00:15:16.020 --> 00:15:20.980
<v Michael Kennedy>the thing could just connect and run database index backed super fast queries,

00:15:21.220 --> 00:15:21.280
<v Michael Kennedy>right?

00:15:21.290 --> 00:15:23.020
<v Michael Kennedy>So it gets really quick answers and answers.

00:15:23.780 --> 00:15:25.500
<v Pamela Fox>I'm seeing now a lot of conferences.

00:15:26.000 --> 00:15:32.600
<v Pamela Fox>Well, not maybe a lot, but the nerdiest conferences will actually have MCP servers to interact with their data, right?

00:15:32.780 --> 00:15:34.640
<v Pamela Fox>So AI engineer did that and that was really fun.

00:15:34.920 --> 00:15:35.380
<v Michael Kennedy>That's awesome.

00:15:35.660 --> 00:15:39.240
<v Michael Kennedy>example for the talk's schedule and stuff.

00:15:39.300 --> 00:15:39.400
<v Michael Kennedy>Like,

00:15:39.800 --> 00:15:41.200
<v Michael Kennedy>hey, I'm interested in this.

00:15:41.260 --> 00:15:42.540
<v Michael Kennedy>What talk should I go see next?

00:15:42.900 --> 00:15:43.000
<v Michael Kennedy>Right.

00:15:43.140 --> 00:15:45.160
<v Michael Kennedy>That kind of question would be awesome to answer right away.

00:15:45.580 --> 00:15:45.680
<v Pamela Fox>Yeah.

00:15:45.720 --> 00:15:45.920
<v Pamela Fox>And then,

00:15:46.140 --> 00:15:50.340
<v Pamela Fox>so here's what's nice about MSP is that then you can connect multiple MSP servers

00:15:50.540 --> 00:15:50.860
<v Pamela Fox>to your agent,

00:15:51.020 --> 00:15:51.080
<v Pamela Fox>right?

00:15:51.120 --> 00:15:51.400
<v Pamela Fox>So example,

00:15:51.600 --> 00:15:52.620
<v Pamela Fox>you could connect the,

00:15:52.900 --> 00:15:57.200
<v Pamela Fox>you know, like the AI engineer MSP server for that conference.

00:15:57.320 --> 00:15:59.400
<v Pamela Fox>And you could also connect like the Google Calendar,

00:15:59.800 --> 00:15:59.900
<v Pamela Fox>like,

00:16:00.100 --> 00:16:02.160
<v Pamela Fox>you know, or in my case, I

00:16:02.020 --> 00:16:05.820
<v Pamela Fox>use the Microsoft WorkIQ MCP server for access to my Outlook.

00:16:06.400 --> 00:16:07.580
<v Pamela Fox>And so then in theory, you can be

00:16:07.660 --> 00:16:07.920
<v Pamela Fox>like, hey,

00:16:08.080 --> 00:16:10.680
<v Pamela Fox>which of these talks am I even like available to attend?

00:16:11.660 --> 00:16:12.740
<v Pamela Fox>Could you also just like

00:16:13.080 --> 00:16:14.040
<v Pamela Fox>put it on my calendar?

00:16:14.520 --> 00:16:14.640
<v Pamela Fox>Right.

00:16:15.320 --> 00:16:20.520
<v Pamela Fox>So this is what excites me about MCP is that MCP enables like

00:16:20.800 --> 00:16:22.680
<v Pamela Fox>us to make mashups really easily.

00:16:22.880 --> 00:16:22.940
<v Pamela Fox>Right.

00:16:23.180 --> 00:16:23.320
<v Pamela Fox>Like,

00:16:23.560 --> 00:16:25.519
<v Pamela Fox>you know, I don't know how people know the term

00:16:25.540 --> 00:16:33.140
<v Pamela Fox>mashups but back with like web 2.0 i love it yeah it was like 2006 like we're you know we had this

00:16:33.400 --> 00:16:41.180
<v Pamela Fox>huge boom of the hdgp web 2.0 apis that were very you know fun to use like the flicker api and google

00:16:41.360 --> 00:16:45.820
<v Pamela Fox>maps and all that stuff and i was making like hundreds of mashups i was absolutely addicted

00:16:46.260 --> 00:16:50.560
<v Pamela Fox>um but it was it was so fun but like i was able to do that because i was you know i was a coder

00:16:50.920 --> 00:16:55.500
<v Pamela Fox>now the difference is you know obviously yes as coders we can make these mashups but we can

00:16:55.520 --> 00:16:58.380
<v Pamela Fox>enable anyone to make these mashups,

00:16:58.560 --> 00:17:01.180
<v Pamela Fox>right? Like we're putting the power of this,

00:17:01.340 --> 00:17:01.560
<v Pamela Fox>you know,

00:17:01.950 --> 00:17:05.579
<v Pamela Fox>this, this programmatic combinatorial power,

00:17:05.730 --> 00:17:07.120
<v Pamela Fox>we're giving it to everyone now.

00:17:07.390 --> 00:17:10.439
<v Michael Kennedy>Yeah. For all the trouble that AI is causing trouble, I don't know,

00:17:10.600 --> 00:17:10.860
<v Michael Kennedy>disruption,

00:17:11.689 --> 00:17:12.100
<v Michael Kennedy>I guess,

00:17:12.260 --> 00:17:13.880
<v Michael Kennedy>in our industry and others,

00:17:14.280 --> 00:17:16.980
<v Michael Kennedy>you know, I do think there's something very special about being able

00:17:17.030 --> 00:17:22.060
<v Michael Kennedy>to give people this power to speak in regular English or whatever language you speak to a

00:17:22.079 --> 00:17:24.100
<v Michael Kennedy>computer and it's almost like programming.

00:17:24.500 --> 00:17:24.560
<v Michael Kennedy>Yeah.

00:17:24.780 --> 00:17:26.699
<v Pamela Fox>Yeah. I mean, I think that's also just generally

00:17:26.959 --> 00:17:27.020
<v Pamela Fox>true,

00:17:27.560 --> 00:17:28.580
<v Pamela Fox>not just with MCP,

00:17:28.800 --> 00:17:29.000
<v Pamela Fox>but,

00:17:29.280 --> 00:17:32.480
<v Pamela Fox>you know, I gave a talk last week to CS teachers talking about,

00:17:32.620 --> 00:17:33.980
<v Pamela Fox>you know, the stack of programming languages.

00:17:34.240 --> 00:17:35.920
<v Pamela Fox>And we went up from, you know, assembly,

00:17:36.340 --> 00:17:36.400
<v Pamela Fox>machine,

00:17:36.840 --> 00:17:37.020
<v Pamela Fox>assembly,

00:17:38.340 --> 00:17:38.540
<v Pamela Fox>you know,

00:17:38.760 --> 00:17:39.160
<v Pamela Fox>C++,

00:17:39.500 --> 00:17:40.320
<v Pamela Fox>higher level like Python.

00:17:40.620 --> 00:17:40.960
<v Pamela Fox>And now,

00:17:41.360 --> 00:17:43.080
<v Pamela Fox>you know, we're at natural human language,

00:17:43.340 --> 00:17:49.620
<v Pamela Fox>right? Like we have lowered the barrier so that anyone spoken language can basically code something,

00:17:50.080 --> 00:17:50.140
<v Pamela Fox>Right.

00:17:50.440 --> 00:17:50.540
<v Pamela Fox>So,

00:17:51.080 --> 00:17:51.200
<v Pamela Fox>yeah.

00:17:51.500 --> 00:17:55.380
<v Michael Kennedy>Yeah. My daughter just discovered Codex and she wanted to build a website.

00:17:55.670 --> 00:17:56.900
<v Michael Kennedy>And after a few hours, she goes,

00:17:57.140 --> 00:17:58.780
<v Michael Kennedy>Dad, is it OK if I stay up all night?

00:17:58.830 --> 00:17:59.640
<v Michael Kennedy>I can't put this down.

00:17:59.860 --> 00:17:59.920
<v Michael Kennedy>Like,

00:18:00.180 --> 00:18:01.980
<v Michael Kennedy>you're going to run out of credits before you stay up all night.

00:18:02.030 --> 00:18:02.720
<v Michael Kennedy>But don't do that.

00:18:03.960 --> 00:18:05.700
<v Michael Kennedy>But that's kind of how the excitement is, you know.

00:18:06.540 --> 00:18:09.140
<v Pamela Fox>That reminds me of discovering the web like in like sixth grade.

00:18:09.210 --> 00:18:10.260
<v Pamela Fox>I definitely stayed up all.

00:18:10.430 --> 00:18:12.480
<v Pamela Fox>The first time you discover that you can make,

00:18:12.780 --> 00:18:15.020
<v Pamela Fox>you know, like a web app or a piece of software,

00:18:15.380 --> 00:18:16.040
<v Pamela Fox>it's a long night.

00:18:17.460 --> 00:18:18.200
<v Michael Kennedy>It's an amazing night.

00:18:18.300 --> 00:18:19.180
<v Michael Kennedy>But the next day, oh,

00:18:19.560 --> 00:18:20.320
<v Michael Kennedy>boy, what have I done?

00:18:20.640 --> 00:18:21.160
<v Michael Kennedy>What have I done?

00:18:21.560 --> 00:18:23.100
<v Michael Kennedy>So make this a little bit concrete.

00:18:23.400 --> 00:18:27.120
<v Michael Kennedy>Give us some specific MCP servers that you use.

00:18:27.820 --> 00:18:30.440
<v Michael Kennedy>And I'll share a couple of mine as well, but I'm going to let you go first on this.

00:18:30.780 --> 00:18:31.040
<v Pamela Fox>Yeah.

00:18:31.220 --> 00:18:33.080
<v Pamela Fox>So the ones that I use the most often,

00:18:33.340 --> 00:18:36.540
<v Pamela Fox>there's the Twitter X MCP server.

00:18:37.380 --> 00:18:38.640
<v Pamela Fox>So I have that installed.

00:18:39.520 --> 00:18:44.980
<v Pamela Fox>And that, you know, can search through all the tweets, especially the tweets I've liked and bookmarked because I use that basically as like research,

00:18:45.260 --> 00:18:45.380
<v Pamela Fox>right?

00:18:45.600 --> 00:18:49.040
<v Pamela Fox>because I obviously am using Twitter purely for educational purposes

00:18:50.180 --> 00:18:51.840
<v Pamela Fox>and, you know, not just for procrastinating.

00:18:52.800 --> 00:18:54.460
<v Pamela Fox>So every week it actually like goes through and,

00:18:54.620 --> 00:18:55.320
<v Pamela Fox>you know, looks at stuff.

00:18:55.470 --> 00:18:58.420
<v Pamela Fox>And it also is like helps me like manage bookmarks,

00:18:58.680 --> 00:18:58.780
<v Pamela Fox>right?

00:18:58.900 --> 00:19:00.480
<v Pamela Fox>So I can put all my bookmarks in one place.

00:19:00.640 --> 00:19:02.140
<v Pamela Fox>So yeah, there's the XMCP server.

00:19:02.340 --> 00:19:03.760
<v Michael Kennedy>I didn't know this existed.

00:19:04.160 --> 00:19:04.340
<v Michael Kennedy>This is,

00:19:04.820 --> 00:19:06.220
<v Michael Kennedy>this could be a problem if I learn about this.

00:19:06.660 --> 00:19:07.020
<v Pamela Fox>Yeah.

00:19:07.860 --> 00:19:08.220
<v Pamela Fox>Yeah,

00:19:08.420 --> 00:19:08.480
<v Pamela Fox>yeah.

00:19:08.960 --> 00:19:10.360
<v Pamela Fox>They've been improving it over time.

00:19:10.390 --> 00:19:12.180
<v Pamela Fox>You still, there's still a fair bit of setup for it.

00:19:12.260 --> 00:19:13.120
<v Pamela Fox>It's not the most,

00:19:13.700 --> 00:19:15.160
<v Pamela Fox>you know, most user friendly.

00:19:15.580 --> 00:19:24.640
<v Pamela Fox>um but uh you know it's it's working decently well uh let's say i also have the um a gmail mcp server

00:19:25.200 --> 00:19:29.640
<v Pamela Fox>which is not quite an official one but it's working well enough for now we're still waiting for it i

00:19:29.740 --> 00:19:33.560
<v Pamela Fox>think they're about to come out with this like google workspace mcp server so i'm excited for

00:19:33.620 --> 00:19:39.320
<v Pamela Fox>that oh that is exciting yeah yeah i basically my friend like vibed an unofficial this is the thing

00:19:39.380 --> 00:19:43.659
<v Pamela Fox>is you can make your own mcp servers that wrap things if you need it so right now i'm using like

00:19:43.680 --> 00:19:45.640
<v Pamela Fox>an official vibe Gmail MCP server.

00:19:46.640 --> 00:19:48.000
<v Pamela Fox>I use the WorkIQ MCP server.

00:19:48.060 --> 00:19:50.620
<v Pamela Fox>That gives me access to my work, Microsoft stuff,

00:19:50.740 --> 00:19:51.200
<v Pamela Fox>like Teams,

00:19:51.400 --> 00:19:51.740
<v Pamela Fox>Outlook,

00:19:52.700 --> 00:19:53.660
<v Pamela Fox>that sort of thing, Calendar.

00:19:54.860 --> 00:19:57.040
<v Pamela Fox>And yeah, and then I have some documentation servers.

00:19:57.160 --> 00:19:59.220
<v Pamela Fox>Like we have a documentation server just for Microsoft Learn,

00:19:59.340 --> 00:20:00.760
<v Pamela Fox>and that's where all of our documentation is.

00:20:01.380 --> 00:20:05.840
<v Pamela Fox>So I'm not even using that many MCP servers,

00:20:07.220 --> 00:20:10.120
<v Pamela Fox>but the ones I am using are just absolutely essential,

00:20:10.340 --> 00:20:10.460
<v Pamela Fox>right?

00:20:10.560 --> 00:20:12.420
<v Pamela Fox>So like every week I hold an office hours

00:20:12.420 --> 00:20:15.060
<v Pamela Fox>where we talk about what's going on in generative AI.

00:20:15.530 --> 00:20:17.360
<v Pamela Fox>And I have a skill for that, an agent skill

00:20:17.520 --> 00:20:20.840
<v Pamela Fox>that specifically uses all those MCP servers to go to research.

00:20:21.080 --> 00:20:22.380
<v Pamela Fox>So it goes and looks through my Gmail,

00:20:22.600 --> 00:20:23.660
<v Pamela Fox>it looks through my Outlook,

00:20:23.690 --> 00:20:26.540
<v Pamela Fox>it looks through my Twitter history and says,

00:20:26.650 --> 00:20:28.500
<v Pamela Fox>okay, these are everything that's happened

00:20:28.860 --> 00:20:30.080
<v Pamela Fox>in the last seven days,

00:20:30.530 --> 00:20:30.660
<v Pamela Fox>right?

00:20:30.960 --> 00:20:33.820
<v Pamela Fox>And it can just do that every Tuesday at 9 a.m.,

00:20:34.020 --> 00:20:34.700
<v Pamela Fox>automatically run,

00:20:35.110 --> 00:20:35.700
<v Pamela Fox>and then boom,

00:20:35.900 --> 00:20:37.100
<v Pamela Fox>got my news roundup.

00:20:37.540 --> 00:20:40.120
<v Pamela Fox>And it's based off the last week

00:20:40.140 --> 00:20:43.280
<v Pamela Fox>of what I have been engaging with myself personally,

00:20:43.600 --> 00:20:43.700
<v Pamela Fox>right?

00:20:44.020 --> 00:20:44.140
<v Michael Kennedy>Yeah,

00:20:44.360 --> 00:20:45.840
<v Michael Kennedy>and for your role, it's really important

00:20:45.940 --> 00:20:47.260
<v Michael Kennedy>to kind of stay on top of the news

00:20:47.440 --> 00:20:49.720
<v Michael Kennedy>and what's happening in tech and all that, right?

00:20:49.940 --> 00:20:51.400
<v Michael Kennedy>So being able to ask that as a question.

00:20:51.780 --> 00:20:51.980
<v Pamela Fox>Right,

00:20:52.160 --> 00:20:52.520
<v Pamela Fox>yeah, that's true.

00:20:52.560 --> 00:20:53.280
<v Pamela Fox>Like for every role,

00:20:53.440 --> 00:20:54.380
<v Pamela Fox>there's different MCP servers

00:20:54.580 --> 00:20:55.200
<v Pamela Fox>that are really shy.

00:20:55.280 --> 00:20:55.940
<v Pamela Fox>Like for other people,

00:20:56.180 --> 00:20:56.380
<v Pamela Fox>they're,

00:20:56.560 --> 00:20:56.680
<v Pamela Fox>you know,

00:20:56.780 --> 00:20:57.940
<v Pamela Fox>like the linear MCP server,

00:20:58.140 --> 00:20:59.140
<v Pamela Fox>they've had a huge,

00:20:59.260 --> 00:21:01.760
<v Pamela Fox>I think they get more traffic from MCP

00:21:02.100 --> 00:21:04.340
<v Pamela Fox>than via like their UI now or something.

00:21:04.460 --> 00:21:06.260
<v Pamela Fox>And they shared some chart where they're like,

00:21:06.600 --> 00:21:08.300
<v Pamela Fox>MCP usage is just, you know,

00:21:08.480 --> 00:21:10.000
<v Pamela Fox>shot way up, right?

00:21:10.080 --> 00:21:11.180
<v Pamela Fox>And that makes a ton of sense, right?

00:21:11.360 --> 00:21:14.020
<v Pamela Fox>Like if you're tracking all your issues in linear,

00:21:15.140 --> 00:21:16.540
<v Pamela Fox>you know, then you're going to have a lot of use.

00:21:17.060 --> 00:21:17.880
<v Pamela Fox>I also obviously use,

00:21:18.100 --> 00:21:18.520
<v Pamela Fox>I'm sorry,

00:21:18.540 --> 00:21:19.940
<v Pamela Fox>I didn't even think about the GitHub MSP server

00:21:20.100 --> 00:21:22.240
<v Pamela Fox>because it's like built in GoPilot,

00:21:22.440 --> 00:21:23.080
<v Pamela Fox>but you know that,

00:21:23.560 --> 00:21:23.660
<v Pamela Fox>you know,

00:21:23.680 --> 00:21:25.000
<v Pamela Fox>because my issues are stored in GitHub.

00:21:25.660 --> 00:21:27.280
<v Pamela Fox>There's Azure DevOps that I'll use sometimes

00:21:27.580 --> 00:21:29.420
<v Pamela Fox>if I'm working on a project that is there.

00:21:29.640 --> 00:21:31.480
<v Pamela Fox>So depending on what your role is,

00:21:31.600 --> 00:21:32.580
<v Pamela Fox>you know, what your job is,

00:21:33.020 --> 00:21:34.540
<v Pamela Fox>you know, there's going to be different MSP servers

00:21:34.760 --> 00:21:36.580
<v Pamela Fox>that are going to like really enhance your productivity

00:21:37.080 --> 00:21:38.300
<v Pamela Fox>and other ones that you just don't need.

00:21:38.500 --> 00:21:38.660
<v Michael Kennedy>Yeah,

00:21:38.940 --> 00:21:39.140
<v Michael Kennedy>very cool.

00:21:39.520 --> 00:21:40.300
<v Michael Kennedy>So for me,

00:21:40.560 --> 00:21:43.780
<v Michael Kennedy>I make a lot of use of the Talk Python one

00:21:43.880 --> 00:21:45.140
<v Michael Kennedy>because it lets me ask questions like,

00:21:45.220 --> 00:21:46.160
<v Michael Kennedy>have we talked about this before?

00:21:46.420 --> 00:21:46.940
<v Michael Kennedy>What have we covered?

00:21:47.300 --> 00:21:50.420
<v Michael Kennedy>Is it going to be just a repeat or that kind of stuff really nicely?

00:21:51.300 --> 00:21:54.820
<v Michael Kennedy>One, just to give a bit of an outside-of-the-box angle,

00:21:55.100 --> 00:21:56.400
<v Michael Kennedy>is I found this place.

00:21:59.020 --> 00:21:59.420
<v Michael Kennedy>Also,

00:21:59.740 --> 00:22:00.340
<v Michael Kennedy>Century MCP.

00:22:00.680 --> 00:22:03.600
<v Michael Kennedy>Century MCP is awesome because I have Century plugged into all my apps,

00:22:03.900 --> 00:22:07.320
<v Michael Kennedy>and I love it for my web apps, but I'm a good web developer.

00:22:07.360 --> 00:22:08.520
<v Michael Kennedy>I can do this, but you know what?

00:22:08.580 --> 00:22:10.620
<v Michael Kennedy>I'm not a good developer of mobile apps.

00:22:10.900 --> 00:22:16.500
<v Michael Kennedy>I'm barely hanging on to the edge to keep my mobile app going because I didn't write it.

00:22:16.530 --> 00:22:17.160
<v Michael Kennedy>I don't know it.

00:22:17.220 --> 00:22:18.260
<v Michael Kennedy>Like, it's just like, but okay,

00:22:18.560 --> 00:22:18.800
<v Michael Kennedy>sort of.

00:22:19.080 --> 00:22:24.500
<v Michael Kennedy>And I had Sentry connected to it for years and it just built up this huge backlog of errors.

00:22:25.000 --> 00:22:25.620
<v Michael Kennedy>And I'm like, you know what?

00:22:25.860 --> 00:22:26.680
<v Michael Kennedy>I'm tired of these errors.

00:22:27.040 --> 00:22:28.340
<v Michael Kennedy>I think one week and I said,

00:22:28.500 --> 00:22:29.820
<v Michael Kennedy>I bet Claude and I,

00:22:30.190 --> 00:22:32.060
<v Michael Kennedy>given all that Sentry data with all the errors,

00:22:32.150 --> 00:22:34.220
<v Michael Kennedy>I bet we can go fix almost all of them.

00:22:34.370 --> 00:22:38.260
<v Michael Kennedy>And so I just connected the MCP server and I opened up VS Code with Flutter and I said,

00:22:38.360 --> 00:22:38.600
<v Michael Kennedy>all right,

00:22:38.860 --> 00:22:39.040
<v Michael Kennedy>Claude,

00:22:39.210 --> 00:22:39.980
<v Michael Kennedy>I want you to look here.

00:22:40.280 --> 00:22:42.620
<v Michael Kennedy>We're going to go there one at a time until these are gone.

00:22:43.400 --> 00:22:44.500
<v Michael Kennedy>And it took three or four days,

00:22:44.740 --> 00:22:47.160
<v Michael Kennedy>but I completely refreshed the app

00:22:47.500 --> 00:22:49.500
<v Michael Kennedy>with all the details from the MCP server.

00:22:49.590 --> 00:22:50.980
<v Michael Kennedy>And I would just say, okay, what's next?

00:22:51.150 --> 00:22:52.080
<v Michael Kennedy>Yeah, go ahead and close that one.

00:22:52.160 --> 00:22:53.240
<v Michael Kennedy>Let's make a GitHub issue for that.

00:22:53.340 --> 00:22:53.680
<v Michael Kennedy>You know what I mean?

00:22:53.880 --> 00:22:54.140
<v Michael Kennedy>And just,

00:22:54.600 --> 00:22:56.660
<v Michael Kennedy>I kind of was just guiding it.

00:22:56.730 --> 00:22:57.580
<v Michael Kennedy>And it completely,

00:22:57.800 --> 00:22:59.000
<v Michael Kennedy>I think this is an amazing thing.

00:22:59.539 --> 00:23:01.960
<v Michael Kennedy>These error-based systems that catch all the errors

00:23:02.010 --> 00:23:02.700
<v Michael Kennedy>and you can just say,

00:23:03.040 --> 00:23:04.120
<v Michael Kennedy>go to the Claude or Codex,

00:23:04.260 --> 00:23:04.320
<v Michael Kennedy>go,

00:23:04.520 --> 00:23:05.860
<v Michael Kennedy>there's an error in Sentry,

00:23:06.190 --> 00:23:06.440
<v Michael Kennedy>fix it.

00:23:06.670 --> 00:23:07.060
<v Michael Kennedy>You know what I mean?

00:23:07.160 --> 00:23:07.720
<v Michael Kennedy>That's all it needs.

00:23:08.060 --> 00:23:08.280
<v Pamela Fox>Yeah,

00:23:08.540 --> 00:23:08.920
<v Pamela Fox>that's amazing.

00:23:09.060 --> 00:23:11.260
<v Pamela Fox>I can remember using Centurate Khan Academy

00:23:11.460 --> 00:23:13.520
<v Pamela Fox>and having those errors just build up over time.

00:23:13.550 --> 00:23:15.680
<v Pamela Fox>And you're like, one day, one day I'm going to fix them.

00:23:15.710 --> 00:23:18.820
<v Pamela Fox>But it is so hard to prioritize bugs

00:23:19.000 --> 00:23:19.960
<v Pamela Fox>when you're always like features.

00:23:20.140 --> 00:23:22.680
<v Pamela Fox>But now we can paralyze a bit more

00:23:23.140 --> 00:23:26.100
<v Pamela Fox>and you get the help of being able to sort through logs.

00:23:26.260 --> 00:23:27.620
<v Pamela Fox>At this point, I'm never going to look through

00:23:27.620 --> 00:23:28.620
<v Pamela Fox>an error traceback myself,

00:23:29.160 --> 00:23:30.240
<v Pamela Fox>especially with Python async.

00:23:30.290 --> 00:23:31.960
<v Pamela Fox>Man, Python async tangent.

00:23:32.200 --> 00:23:33.720
<v Pamela Fox>But Python async errors,

00:23:33.900 --> 00:23:34.920
<v Pamela Fox>the tracebacks

00:23:35.050 --> 00:23:37.540
<v Pamela Fox>are so hard to parse through, right?

00:23:37.940 --> 00:23:43.940
<v Pamela Fox>So these days I'm just always going to paste a trace back into an agent and be like, okay, what's the actual error here,

00:23:44.120 --> 00:23:44.180
<v Pamela Fox>right?

00:23:44.440 --> 00:23:44.500
<v Michael Kennedy>Yeah.

00:23:44.900 --> 00:23:45.600
<v Michael Kennedy>This is wrong.

00:23:46.760 --> 00:23:47.680
<v Michael Kennedy>There's something wrong.

00:23:47.760 --> 00:23:48.520
<v Michael Kennedy>I know this is wrong.

00:23:48.720 --> 00:23:49.620
<v Michael Kennedy>We're fixing this,

00:23:49.800 --> 00:23:49.900
<v Michael Kennedy>right?

00:23:50.160 --> 00:23:53.920
<v Michael Kennedy>And then another one that I think we'll see if you would see this coming.

00:23:54.250 --> 00:23:55.240
<v Michael Kennedy>So there's this place.

00:23:55.510 --> 00:23:57.480
<v Michael Kennedy>So I was thinking about like retirement planning, right?

00:23:57.640 --> 00:23:59.340
<v Michael Kennedy>Eventually I'm going to retire someday.

00:23:59.710 --> 00:24:02.460
<v Michael Kennedy>And I'm like, I ran across this place called Bolden.

00:24:02.540 --> 00:24:03.320
<v Michael Kennedy>It's not an advertisement.

00:24:03.480 --> 00:24:04.420
<v Michael Kennedy>Just thought it was interesting.

00:24:04.820 --> 00:24:05.300
<v Michael Kennedy>And it says,

00:24:05.500 --> 00:24:07.080
<v Michael Kennedy>it's the DIY retirement.

00:24:07.380 --> 00:24:09.580
<v Michael Kennedy>Like you can run any what if on their AI.

00:24:10.160 --> 00:24:16.180
<v Michael Kennedy>If you put your sort of financial situation into it, you can ask, like, if I put more in 401k now,

00:24:16.460 --> 00:24:16.880
<v Michael Kennedy>will that save?

00:24:17.100 --> 00:24:19.260
<v Michael Kennedy>You know, like what's this going to have an effect on, right?

00:24:19.340 --> 00:24:20.260
<v Michael Kennedy>And if the market does this,

00:24:20.420 --> 00:24:20.940
<v Michael Kennedy>what would that do?

00:24:20.980 --> 00:24:22.040
<v Michael Kennedy>You know, like these kinds of things.

00:24:22.420 --> 00:24:25.300
<v Michael Kennedy>And I asked it a question and I was like, I'll try this AI thing.

00:24:25.640 --> 00:24:26.420
<v Michael Kennedy>So I asked it a question.

00:24:26.480 --> 00:24:27.960
<v Michael Kennedy>It came back like that with an answer.

00:24:27.960 --> 00:24:28.340
<v Michael Kennedy>I'm like, oh,

00:24:28.540 --> 00:24:29.660
<v Michael Kennedy>no, I don't trust this at all.

00:24:29.960 --> 00:24:31.580
<v Michael Kennedy>This should have taken a little bit of thinking.

00:24:31.760 --> 00:24:34.320
<v Michael Kennedy>This is like a haiku model at best.

00:24:34.560 --> 00:24:34.940
<v Michael Kennedy>You know what I mean?

00:24:35.120 --> 00:24:36.500
<v Michael Kennedy>There's no way there's a good AI here.

00:24:36.860 --> 00:24:37.840
<v Michael Kennedy>So I complained to them.

00:24:38.200 --> 00:24:41.240
<v Michael Kennedy>And the developer wrote back and said, well, what if we build an MCP server?

00:24:41.320 --> 00:24:41.640
<v Michael Kennedy>I said,

00:24:41.800 --> 00:24:41.960
<v Michael Kennedy>yes.

00:24:42.340 --> 00:24:44.320
<v Michael Kennedy>And then a few weeks later, like, here's the MCP server.

00:24:44.720 --> 00:24:48.260
<v Michael Kennedy>So I connected it to Claude on like Fable or Opus or something.

00:24:48.440 --> 00:24:49.300
<v Michael Kennedy>And then I'm like, okay,

00:24:49.520 --> 00:24:51.760
<v Michael Kennedy>now I can ask questions I trust the answers to.

00:24:52.160 --> 00:24:52.920
<v Pamela Fox>Yeah, that's a good point.

00:24:53.020 --> 00:24:53.700
<v Pamela Fox>Because when you,

00:24:54.200 --> 00:24:54.840
<v Pamela Fox>right, that's a good point.

00:24:54.920 --> 00:24:57.300
<v Pamela Fox>Because a lot of these applications have,

00:24:57.540 --> 00:24:58.520
<v Pamela Fox>you know, this has like an agent.

00:24:58.640 --> 00:24:59.500
<v Pamela Fox>You can see it on this,

00:24:59.680 --> 00:25:00.920
<v Pamela Fox>the front web page.

00:25:00.920 --> 00:25:02.320
<v Pamela Fox>It says like AI planner assistant.

00:25:02.640 --> 00:25:06.800
<v Pamela Fox>But they need to optimize for cost, so they're not necessarily going to use the most expensive model there.

00:25:06.870 --> 00:25:08.720
<v Pamela Fox>They may not have access to the best model.

00:25:09.060 --> 00:25:10.420
<v Pamela Fox>You might,

00:25:10.620 --> 00:25:12.160
<v Pamela Fox>for whatever reason, you've got Fable.

00:25:12.390 --> 00:25:14.720
<v Pamela Fox>I don't even have Fable, but I've got Opus 5.5,

00:25:14.960 --> 00:25:15.960
<v Pamela Fox>which is my new bestie, right?

00:25:16.420 --> 00:25:17.220
<v Michael Kennedy>I love 5.5.

00:25:17.270 --> 00:25:18.240
<v Michael Kennedy>Yeah, I love Opus 5.5.

00:25:18.400 --> 00:25:19.020
<v Michael Kennedy>It's so good.

00:25:19.220 --> 00:25:20.140
<v Pamela Fox>So I can.55,

00:25:20.920 --> 00:25:23.840
<v Pamela Fox>and I can trust that.

00:25:24.780 --> 00:25:28.500
<v Pamela Fox>And we should be using the EPSP server for the data.

00:25:29.240 --> 00:25:32.580
<v Pamela Fox>There are some MSP servers that do also basically wrap an agent.

00:25:32.970 --> 00:25:35.240
<v Pamela Fox>I don't love that as a design choice,

00:25:35.720 --> 00:25:40.320
<v Pamela Fox>but it is certainly a choice that some MSP servers are using.

00:25:41.540 --> 00:25:44.460
<v Pamela Fox>But yeah, I'd rather that they focus on returning the data

00:25:44.760 --> 00:25:47.360
<v Pamela Fox>and leave the agentic processing to people's agents.

00:25:47.820 --> 00:25:47.940
<v Michael Kennedy>Right.

00:25:48.840 --> 00:25:51.880
<v Michael Kennedy>Your job is to get the data to the smart agent that I'm running,

00:25:51.980 --> 00:25:53.420
<v Michael Kennedy>not do your own thing on top of it.

00:25:53.450 --> 00:25:53.580
<v Michael Kennedy>Exactly.

00:25:54.680 --> 00:25:56.480
<v Michael Kennedy>I want to throw one more thing out there really quick,

00:25:56.490 --> 00:25:59.040
<v Michael Kennedy>and maybe you can give us your thoughts if you have any.

00:25:59.300 --> 00:26:03.260
<v Michael Kennedy>I know a lot of people when they think about AI and their website or their app,

00:26:03.420 --> 00:26:05.680
<v Michael Kennedy>they're like, here's a thing that is swamping my app.

00:26:05.860 --> 00:26:06.780
<v Michael Kennedy>It's making it slow.

00:26:07.040 --> 00:26:08.000
<v Michael Kennedy>It's overwhelming the bandwidth.

00:26:08.200 --> 00:26:08.820
<v Michael Kennedy>And I hate it.

00:26:09.120 --> 00:26:10.760
<v Michael Kennedy>So MCP probably just makes this worse.

00:26:11.700 --> 00:26:12.940
<v Michael Kennedy>I don't think that's true.

00:26:13.200 --> 00:26:16.040
<v Michael Kennedy>I think that a lot of people might have this sort of thought process,

00:26:16.220 --> 00:26:16.340
<v Michael Kennedy>right?

00:26:16.500 --> 00:26:19.780
<v Michael Kennedy>Like it's just going to mean it's going to talk to my site more and so on.

00:26:20.060 --> 00:26:22.360
<v Michael Kennedy>But when you build these MCP tools,

00:26:22.520 --> 00:26:24.340
<v Michael Kennedy>as you pointed out, you build the tools.

00:26:24.540 --> 00:26:26.820
<v Michael Kennedy>Those are basically function calls on your code.

00:26:27.160 --> 00:26:28.880
<v Michael Kennedy>And you can do whatever you want there.

00:26:29.020 --> 00:26:29.760
<v Michael Kennedy>You can use caching.

00:26:30.030 --> 00:26:31.460
<v Michael Kennedy>You can use database indexes.

00:26:31.530 --> 00:26:32.840
<v Michael Kennedy>You can make them incredibly fast.

00:26:33.150 --> 00:26:35.560
<v Michael Kennedy>So instead of having an AI crawl all over your stuff,

00:26:35.980 --> 00:26:38.000
<v Michael Kennedy>it can just go right to an optimized spot and go,

00:26:38.380 --> 00:26:39.360
<v Michael Kennedy>this is what I'm actually looking for.

00:26:39.520 --> 00:26:41.600
<v Michael Kennedy>All right, here's a super fast response.

00:26:42.540 --> 00:26:43.780
<v Michael Kennedy>Harkening back to the PyCon,

00:26:44.040 --> 00:26:44.480
<v Michael Kennedy>PyBay talk.

00:26:44.760 --> 00:26:48.560
<v Michael Kennedy>Pulled up my logs in the last couple of MCP calls on my site

00:26:48.570 --> 00:26:50.020
<v Michael Kennedy>were like two millisecond response time.

00:26:50.320 --> 00:26:51.160
<v Michael Kennedy>That doesn't hurt my site.

00:26:51.250 --> 00:26:51.680
<v Michael Kennedy>I don't care.

00:26:51.970 --> 00:26:53.240
<v Michael Kennedy>They can ask questions like that.

00:26:53.600 --> 00:26:57.120
<v Michael Kennedy>This is the way to, I think, lessen the work on your server

00:26:57.260 --> 00:26:58.420
<v Michael Kennedy>and your app, not increase it.

00:26:58.640 --> 00:26:58.720
<v Pamela Fox>Yeah,

00:26:58.900 --> 00:26:59.400
<v Pamela Fox>that's really true.

00:26:59.620 --> 00:27:04.600
<v Pamela Fox>I think at this point, we should assume that people are going to use agents to try to automate

00:27:05.419 --> 00:27:07.260
<v Pamela Fox>interaction with the websites they're using,

00:27:07.740 --> 00:27:07.840
<v Pamela Fox>right?

00:27:07.960 --> 00:27:08.520
<v Pamela Fox>So for example,

00:27:08.680 --> 00:27:12.940
<v Pamela Fox>I was trying to programmatically browse Etsy the other day to help with my

00:27:13.180 --> 00:27:18.200
<v Pamela Fox>costume shopping because I wanted to be Link from Zelda and I needed to put together lots

00:27:18.240 --> 00:27:18.800
<v Pamela Fox>of different parts.

00:27:19.680 --> 00:27:20.900
<v Pamela Fox>And I actually have never played Zelda.

00:27:21.320 --> 00:27:21.840
<v Pamela Fox>So I was like,

00:27:22.360 --> 00:27:25.620
<v Pamela Fox>assuming the coding agent knows more about Link than I do,

00:27:25.820 --> 00:27:27.220
<v Pamela Fox>you know, because

00:27:27.320 --> 00:27:27.900
<v Pamela Fox>it's in the weights.

00:27:28.300 --> 00:27:29.220
<v Pamela Fox>It's not in my weights,

00:27:29.480 --> 00:27:29.580
<v Pamela Fox>right?

00:27:29.960 --> 00:27:32.000
<v Pamela Fox>But I was immediately blocked from Etsy.

00:27:32.260 --> 00:27:35.900
<v Pamela Fox>And Etsy does not have like a shopping MCP server.

00:27:36.280 --> 00:27:38.660
<v Pamela Fox>So if they'd had an MCP server,

00:27:38.900 --> 00:27:41.960
<v Pamela Fox>then I would have bought a lot more on Etsy.

00:27:41.960 --> 00:27:42.860
<v Michael Kennedy>Yeah, isn't that surprising?

00:27:43.100 --> 00:27:44.780
<v Michael Kennedy>They're like, no, it's going to block it.

00:27:44.840 --> 00:27:45.240
<v Michael Kennedy>These are terrible.

00:27:45.400 --> 00:27:45.460
<v Michael Kennedy>Like,

00:27:45.680 --> 00:27:47.820
<v Michael Kennedy>I was trying to buy five different things for an outfit.

00:27:48.020 --> 00:27:48.620
<v Michael Kennedy>You wouldn't let me,

00:27:48.920 --> 00:27:49.020
<v Michael Kennedy>basically.

00:27:49.040 --> 00:27:49.900
<v Pamela Fox>Right, because if you, yeah.

00:27:50.000 --> 00:27:52.020
<v Pamela Fox>So it certainly does depend on what people.

00:27:52.040 --> 00:27:53.460
<v Pamela Fox>So because on the flip side,

00:27:53.840 --> 00:27:57.019
<v Pamela Fox>you know, when people use MCP servers to automate interaction

00:27:57.040 --> 00:27:58.820
<v Pamela Fox>with like a social media network or,

00:27:59.220 --> 00:27:59.660
<v Pamela Fox>you know, like,

00:28:00.140 --> 00:28:02.240
<v Pamela Fox>you know, like a website where people are

00:28:02.500 --> 00:28:04.140
<v Pamela Fox>posting text,

00:28:04.540 --> 00:28:05.680
<v Pamela Fox>they can make a worse experience,

00:28:05.880 --> 00:28:07.840
<v Pamela Fox>right? We see this on Twitter right now with all

00:28:07.960 --> 00:28:08.900
<v Pamela Fox>the reply bots.

00:28:09.720 --> 00:28:09.840
<v Pamela Fox>You know,

00:28:10.000 --> 00:28:10.180
<v Pamela Fox>they're,

00:28:10.360 --> 00:28:10.620
<v Pamela Fox>they're,

00:28:11.020 --> 00:28:11.980
<v Pamela Fox>it's a horrible experience.

00:28:12.200 --> 00:28:13.540
<v Pamela Fox>Like everything you tweet,

00:28:13.980 --> 00:28:17.160
<v Pamela Fox>you get back this reply that's very well written and says it's not X, but Y.

00:28:17.400 --> 00:28:17.660
<v Pamela Fox>And you're like,

00:28:17.800 --> 00:28:17.940
<v Pamela Fox>oh,

00:28:18.200 --> 00:28:18.340
<v Pamela Fox>why?

00:28:18.520 --> 00:28:18.980
<v Pamela Fox>Why did you,

00:28:19.560 --> 00:28:19.720
<v Pamela Fox>you know,

00:28:20.200 --> 00:28:23.379
<v Pamela Fox>like fill my brain with something that I should never have to read

00:28:23.400 --> 00:28:23.620
<v Pamela Fox>because,

00:28:24.180 --> 00:28:25.940
<v Pamela Fox>you know, there's no human out there that wrote it.

00:28:26.140 --> 00:28:26.200
<v Pamela Fox>Right.

00:28:26.640 --> 00:28:26.720
<v Pamela Fox>So,

00:28:27.000 --> 00:28:27.340
<v Pamela Fox>you know, there,

00:28:27.670 --> 00:28:27.880
<v Pamela Fox>you know,

00:28:27.970 --> 00:28:30.980
<v Pamela Fox>there, there's some design that goes into this,

00:28:31.140 --> 00:28:33.840
<v Pamela Fox>but there's definitely many services where if you

00:28:34.070 --> 00:28:34.900
<v Pamela Fox>build an MSP server,

00:28:35.310 --> 00:28:38.880
<v Pamela Fox>you are enabling people to use your app in,

00:28:39.600 --> 00:28:40.440
<v Pamela Fox>in more ways, right? So,

00:28:40.640 --> 00:28:41.220
<v Pamela Fox>especially for shopping,

00:28:41.420 --> 00:28:44.680
<v Pamela Fox>right? So DoorDash, they did add an MSP server and that makes so much

00:28:44.800 --> 00:28:45.100
<v Pamela Fox>sense, right?

00:28:45.260 --> 00:28:45.920
<v Pamela Fox>For people who's doing,

00:28:46.080 --> 00:28:48.060
<v Pamela Fox>you know, maybe they use DoorDash to cater to their office

00:28:48.190 --> 00:28:48.760
<v Pamela Fox>every single day.

00:28:49.300 --> 00:28:50.440
<v Pamela Fox>By having an MSP server,

00:28:50.700 --> 00:28:53.360
<v Pamela Fox>DoorDash is making it that much easier to do bulk orders

00:28:53.380 --> 00:28:54.060
<v Pamela Fox>on DoorDash,

00:28:54.220 --> 00:28:54.340
<v Pamela Fox>right?

00:28:54.650 --> 00:28:55.940
<v Pamela Fox>So if you are a shopping website,

00:28:56.630 --> 00:28:57.880
<v Pamela Fox>I think Shopify just announced

00:28:58.030 --> 00:28:59.900
<v Pamela Fox>like an MCP sort of integration.

00:29:00.160 --> 00:29:01.600
<v Pamela Fox>So if you are a shopping website,

00:29:01.770 --> 00:29:03.100
<v Pamela Fox>it makes so much sense

00:29:03.500 --> 00:29:04.320
<v Pamela Fox>to have an MCP server,

00:29:04.520 --> 00:29:04.580
<v Pamela Fox>right?

00:29:04.630 --> 00:29:05.340
<v Pamela Fox>Because you want that,

00:29:05.570 --> 00:29:05.940
<v Pamela Fox>you know, sweet,

00:29:06.100 --> 00:29:06.880
<v Pamela Fox>sweet consumerism,

00:29:07.140 --> 00:29:07.260
<v Pamela Fox>right?

00:29:07.660 --> 00:29:07.880
<v Michael Kennedy>Exactly.

00:29:08.220 --> 00:29:09.100
<v Michael Kennedy>Let them come purchase.

00:29:09.270 --> 00:29:10.040
<v Michael Kennedy>Let their agents come.

00:29:10.480 --> 00:29:11.360
<v Michael Kennedy>They'll put their credit card in.

00:29:11.580 --> 00:29:12.680
<v Pamela Fox>I guess the thing that it threatens

00:29:12.770 --> 00:29:13.740
<v Pamela Fox>the most is advertising,

00:29:13.960 --> 00:29:14.040
<v Pamela Fox>right?

00:29:14.100 --> 00:29:15.200
<v Pamela Fox>And that's always what's been threatened

00:29:15.720 --> 00:29:16.080
<v Pamela Fox>by,

00:29:16.530 --> 00:29:17.140
<v Pamela Fox>you know, ad block

00:29:17.250 --> 00:29:18.120
<v Pamela Fox>and programmatic scraping

00:29:18.280 --> 00:29:19.000
<v Pamela Fox>and all that stuff.

00:29:19.220 --> 00:29:20.320
<v Pamela Fox>So we'll have to figure that out.

00:29:20.620 --> 00:29:20.760
<v Michael Kennedy>Yeah,

00:29:21.000 --> 00:29:22.020
<v Michael Kennedy>that is a different,

00:29:22.110 --> 00:29:23.340
<v Michael Kennedy>I think that is an absolutely

00:29:23.500 --> 00:29:27.160
<v Michael Kennedy>significant area and totally different though than like etsy letting you buy or not buy

00:29:27.500 --> 00:29:32.120
<v Michael Kennedy>so you gave a talk exploring mcp servers with GitHub copilot was this last year

00:29:32.580 --> 00:29:37.720
<v Pamela Fox>um this particular one was quite uh recent but basically like i keep doing these workshops to

00:29:37.860 --> 00:29:43.060
<v Pamela Fox>show people mcp uh the first one i gave was at pycon so that was you know pycon back down in long

00:29:43.220 --> 00:29:46.820
<v Pamela Fox>beach and that was you know like a three-hour workshop and so i've just been giving variants

00:29:47.160 --> 00:29:53.320
<v Pamela Fox>on this workshop for the last whatever seven months i think this might have been pi ai is that right

00:29:53.340 --> 00:29:58.720
<v Pamela Fox>remember uh maybe not no no no that was your um that was your one where you made an mcp server

00:29:58.850 --> 00:30:03.060
<v Pamela Fox>pick your outfit that was a different deal yeah you know i just talk about mcp everywhere so this

00:30:03.060 --> 00:30:08.220
<v Pamela Fox>was at we are developers in san jose a few weeks ago gotcha okay the reason is i want to talk about

00:30:08.360 --> 00:30:15.580
<v Michael Kennedy>just like support for mcp servers across some of your common things like copilot codex code vs code

00:30:15.860 --> 00:30:21.380
<v Michael Kennedy>etc it's like a can i use but can i use dot ai instead of an air.com whatever it is i know i don't

00:30:21.340 --> 00:30:25.060
<v Michael Kennedy>even think don't please don't go there i don't know what's there but no the actual website is

00:30:25.160 --> 00:30:30.320
<v Pamela Fox>can i use.dev uh i told them they should get canimcp.com because i thought that would you know

00:30:30.550 --> 00:30:35.940
<v Pamela Fox>flow better but this is from so the people who made this is mcp jam it's this really nice uh

00:30:36.160 --> 00:30:40.620
<v Pamela Fox>basically like think of it as postman but for mcp servers right so really nice way of just testing

00:30:40.800 --> 00:30:44.520
<v Pamela Fox>out your mcp servers with all the different features and you know sending requests back and

00:30:44.650 --> 00:30:49.700
<v Pamela Fox>forth and being able to see your oaf flow and see it succeed so they decided to make this can i use.dev

00:30:49.980 --> 00:30:52.240
<v Pamela Fox>And it's supposed to be like caniuse.com,

00:30:52.640 --> 00:30:53.460
<v Pamela Fox>but for MSP servers,

00:30:53.820 --> 00:30:55.560
<v Pamela Fox>because people are always asking like, hey,

00:30:55.820 --> 00:30:56.980
<v Pamela Fox>you know, this new MSP feature,

00:30:57.500 --> 00:30:58.500
<v Pamela Fox>is it supporting these clients?

00:30:58.640 --> 00:31:00.720
<v Pamela Fox>Because this MSP spec is evolving all the time.

00:31:00.760 --> 00:31:04.840
<v Pamela Fox>They did the most recent big drop was July 28th, continually adding,

00:31:05.200 --> 00:31:06.880
<v Pamela Fox>you know, more stuff to it over time.

00:31:07.760 --> 00:31:12.020
<v Pamela Fox>But the problem is many clients don't necessarily immediately add support for these features.

00:31:12.180 --> 00:31:12.640
<v Pamela Fox>So as somebody,

00:31:12.960 --> 00:31:14.040
<v Pamela Fox>if you're building a server,

00:31:14.680 --> 00:31:14.800
<v Pamela Fox>you know,

00:31:14.980 --> 00:31:16.520
<v Pamela Fox>you might be really excited about a new feature,

00:31:17.040 --> 00:31:18.420
<v Pamela Fox>but you want to check and see like, hey,

00:31:18.940 --> 00:31:19.140
<v Pamela Fox>you know,

00:31:19.620 --> 00:31:21.340
<v Pamela Fox>what is the support like that,

00:31:21.600 --> 00:31:21.720
<v Pamela Fox>you know,

00:31:22.160 --> 00:31:23.680
<v Pamela Fox>for this feature across the clients?

00:31:23.800 --> 00:31:25.800
<v Pamela Fox>What is the user experience going to be like?

00:31:26.080 --> 00:31:26.260
<v Michael Kennedy>Yeah.

00:31:26.320 --> 00:31:27.960
<v Michael Kennedy>And just doesn't support it at all.

00:31:28.360 --> 00:31:28.500
<v Michael Kennedy>You know,

00:31:28.660 --> 00:31:29.480
<v Michael Kennedy>I know for a while,

00:31:29.780 --> 00:31:30.480
<v Michael Kennedy>like ChatGPT,

00:31:30.840 --> 00:31:34.760
<v Michael Kennedy>Codex supported, but I think ChatGPT wouldn't take MCP or something.

00:31:35.780 --> 00:31:36.680
<v Michael Kennedy>Why won't you accept it?

00:31:36.740 --> 00:31:38.020
<v Michael Kennedy>You're the most popular thing.

00:31:38.720 --> 00:31:39.160
<v Michael Kennedy>You know what I mean?

00:31:39.200 --> 00:31:41.120
<v Michael Kennedy>I think MCP came out of Anthropic.

00:31:41.220 --> 00:31:41.980
<v Michael Kennedy>I can't remember exactly.

00:31:42.060 --> 00:31:47.380
<v Michael Kennedy>But actually Codex has more support than log code, for example.

00:31:47.530 --> 00:31:49.700
<v Michael Kennedy>So this is kind of a can I use for all the different features.

00:31:51.100 --> 00:31:51.180
<v Pamela Fox>Yeah,

00:31:51.340 --> 00:31:51.780
<v Pamela Fox>and I actually,

00:31:52.160 --> 00:31:53.140
<v Pamela Fox>relevant to this audience,

00:31:53.370 --> 00:31:57.920
<v Pamela Fox>I just asked them if they will add columns for Python agent frameworks,

00:31:58.220 --> 00:31:58.360
<v Pamela Fox>right?

00:31:58.540 --> 00:31:58.660
<v Pamela Fox>Because,

00:31:58.930 --> 00:32:00.420
<v Pamela Fox>for example, skills over MCP,

00:32:00.640 --> 00:32:02.480
<v Pamela Fox>that just got approved like a month ago.

00:32:02.930 --> 00:32:05.300
<v Pamela Fox>Like the spec for it got approved in merge.

00:32:05.330 --> 00:32:06.060
<v Pamela Fox>And so now they're,

00:32:06.220 --> 00:32:06.400
<v Pamela Fox>you know,

00:32:06.580 --> 00:32:07.180
<v Pamela Fox>making it official.

00:32:07.760 --> 00:32:14.700
<v Pamela Fox>And if you look for skills on can I use.dev right now, you'll see none of the current clients support it,

00:32:14.920 --> 00:32:17.200
<v Pamela Fox>but it is supported in a lot of the agent frameworks.

00:32:17.500 --> 00:32:22.360
<v Pamela Fox>So if you use like we've got like Microsoft agent framework, I happen to just know that that one supports it.

00:32:22.360 --> 00:32:23.700
<v Pamela Fox>But I bet you Pydantic AI does,

00:32:23.940 --> 00:32:26.100
<v Pamela Fox>too, because they're always quick to pick these things up.

00:32:26.220 --> 00:32:26.320
<v Pamela Fox>Right.

00:32:26.500 --> 00:32:30.860
<v Pamela Fox>So I did ask them to add columns for the agent framework.

00:32:30.980 --> 00:32:32.240
<v Pamela Fox>So hopefully in the near future,

00:32:32.480 --> 00:32:33.100
<v Pamela Fox>you'll see like,

00:32:33.860 --> 00:32:34.000
<v Pamela Fox>you know,

00:32:34.160 --> 00:32:35.300
<v Pamela Fox>does Pydantic AI,

00:32:35.720 --> 00:32:36.580
<v Pamela Fox>LangChain, right?

00:32:37.000 --> 00:32:39.700
<v Pamela Fox>You know, which of your favorite Python agent framework support it?

00:32:39.700 --> 00:32:41.420
<v Pamela Fox>Because that's the thing is like there's both,

00:32:42.040 --> 00:32:43.360
<v Pamela Fox>you know, using MCP from,

00:32:44.000 --> 00:32:45.740
<v Pamela Fox>you know, almost say like a user facing tool.

00:32:45.920 --> 00:32:50.380
<v Pamela Fox>Because in some ways, like GitHub Copilot is it's like a kind of a user facing tool, even if the users are developers.

00:32:51.880 --> 00:32:57.720
<v Pamela Fox>Versus using MCP as the integration layer for agents that you're writing in Python.

00:32:57.820 --> 00:33:02.260
<v Pamela Fox>And that's actually where we see a huge amount of uses of MCP,

00:33:02.510 --> 00:33:02.640
<v Pamela Fox>right?

00:33:02.710 --> 00:33:04.820
<v Pamela Fox>Because there's always this debate about MCP versus CLI.

00:33:05.110 --> 00:33:05.340
<v Pamela Fox>And yeah,

00:33:05.580 --> 00:33:07.480
<v Pamela Fox>locally, you can do whatever the heck you want,

00:33:07.660 --> 00:33:08.340
<v Pamela Fox>whatever works for you.

00:33:09.420 --> 00:33:19.700
<v Pamela Fox>But when it comes to putting agents into production and having them do communication with other servers when they're actually deployed to the cloud,

00:33:19.880 --> 00:33:21.320
<v Pamela Fox>everyone's using MCP there,

00:33:21.680 --> 00:33:21.800
<v Pamela Fox>right?

00:33:21.880 --> 00:33:24.360
<v Pamela Fox>Like for enterprise agent interconnectedness,

00:33:24.560 --> 00:33:25.580
<v Pamela Fox>it's all MCP.

00:33:25.820 --> 00:33:34.600
<v Pamela Fox>So what I tend to care most about is actually like, hey, can I write a Python agent framework that integrates with a Python agent that integrates with this MCP server?

00:33:34.820 --> 00:33:36.120
<v Pamela Fox>And is it going to work as expected?

00:33:36.460 --> 00:33:39.420
<v Michael Kennedy>So people are excited about MCP probably like, oh,

00:33:39.580 --> 00:33:42.920
<v Michael Kennedy>we're going to add this MCP thing to our topic.

00:33:43.140 --> 00:33:44.420
<v Michael Kennedy>But I think a good question.

00:33:44.780 --> 00:33:45.280
<v Michael Kennedy>And this one,

00:33:45.620 --> 00:33:48.840
<v Michael Kennedy>everything we talk about, we've got to give it a little bit of a caveat of like, well,

00:33:49.180 --> 00:33:53.040
<v Michael Kennedy>was it like last century effectively for this technology like last year?

00:33:53.600 --> 00:33:57.620
<v Michael Kennedy>But you wrote about to MCP or not to MCP, you know,

00:33:57.780 --> 00:33:58.540
<v Michael Kennedy>my weird joke.

00:33:58.700 --> 00:34:02.660
<v Michael Kennedy>But there are reasons to not use it or not completely just leave it wide open,

00:34:02.900 --> 00:34:06.540
<v Michael Kennedy>like create maybe more focus tools and then give those to your AI or to you.

00:34:06.780 --> 00:34:07.000
<v Pamela Fox>Yeah.

00:34:07.340 --> 00:34:08.139
<v Pamela Fox>Yeah. And so here,

00:34:08.720 --> 00:34:11.419
<v Pamela Fox>this was an example for this was Playwright,

00:34:11.679 --> 00:34:11.800
<v Pamela Fox>right?

00:34:12.560 --> 00:34:18.440
<v Pamela Fox>So Playwright is a browser automation library and there's a Python package for it.

00:34:18.560 --> 00:34:19.440
<v Pamela Fox>It's fantastic.

00:34:20.320 --> 00:34:25.260
<v Pamela Fox>And I was already using Playwright quite heavily to write my end-to-end tests for web apps, right?

00:34:25.440 --> 00:34:26.100
<v Pamela Fox>And so you could actually,

00:34:26.260 --> 00:34:28.080
<v Pamela Fox>and there's a pytest-Playwright plugin.

00:34:28.659 --> 00:34:29.700
<v Pamela Fox>So it's very easy to,

00:34:29.899 --> 00:34:32.720
<v Pamela Fox>you know, write these end-to-end tests with Playwright.

00:34:32.840 --> 00:34:36.820
<v Pamela Fox>And they'll, like, automatically open a headless Chromium browser or Firefox browser.

00:34:36.879 --> 00:34:37.639
<v Pamela Fox>And they'll click around.

00:34:37.740 --> 00:34:38.300
<v Pamela Fox>They'll do something.

00:34:38.379 --> 00:34:39.600
<v Pamela Fox>And they'll verify it all works,

00:34:39.800 --> 00:34:39.899
<v Pamela Fox>right?

00:34:40.639 --> 00:34:43.820
<v Pamela Fox>Now, Playwright was one of the first MCP servers that came out.

00:34:44.560 --> 00:34:48.300
<v Pamela Fox>And people loved it because suddenly people discovered browser animation.

00:34:48.520 --> 00:34:49.860
<v Pamela Fox>Like a lot of people, I think, had never,

00:34:50.460 --> 00:34:53.280
<v Pamela Fox>never knew that you could like programmatically control a browser.

00:34:53.429 --> 00:34:54.639
<v Pamela Fox>And so the Playwright MCP,

00:34:54.800 --> 00:34:54.919
<v Pamela Fox>like,

00:34:55.179 --> 00:35:00.720
<v Pamela Fox>you know, it lowered the barrier for that where people were just like adding Playwright MCP server to their coding agents and,

00:35:01.020 --> 00:35:01.940
<v Pamela Fox>you know, having fun,

00:35:02.200 --> 00:35:02.480
<v Pamela Fox>you know,

00:35:02.760 --> 00:35:04.260
<v Pamela Fox>automatically shopping and all that stuff.

00:35:05.260 --> 00:35:05.620
<v Pamela Fox>So,

00:35:06.040 --> 00:35:07.700
<v Pamela Fox>you know, so I that that was, you know,

00:35:08.140 --> 00:35:10.720
<v Pamela Fox>one of the first big servers that you would hear a lot about.

00:35:11.240 --> 00:35:18.240
<v Pamela Fox>So I wanted to make an agent that would automatically triage my LinkedIn request to connect.

00:35:18.640 --> 00:35:19.780
<v Pamela Fox>So you go to a conference,

00:35:20.040 --> 00:35:22.460
<v Pamela Fox>you go back and you see your LinkedIn and all these people

00:35:22.540 --> 00:35:24.780
<v Pamela Fox>want to connect and they're all great people.

00:35:25.240 --> 00:35:27.060
<v Pamela Fox>But I find it like really boring to accept,

00:35:27.400 --> 00:35:27.520
<v Pamela Fox>accept.

00:35:27.900 --> 00:35:29.440
<v Michael Kennedy>And every now and then it's careful,

00:35:29.660 --> 00:35:31.480
<v Michael Kennedy>only connect with people you actually know.

00:35:31.580 --> 00:35:31.940
<v Michael Kennedy>It's like, nope,

00:35:32.220 --> 00:35:32.960
<v Michael Kennedy>accept, accept, accept.

00:35:33.640 --> 00:35:35.120
<v Michael Kennedy>What am I going to be rude and tell them no?

00:35:35.300 --> 00:35:35.980
<v Michael Kennedy>I don't know.

00:35:36.500 --> 00:35:36.580
<v Pamela Fox>Well,

00:35:36.760 --> 00:35:40.500
<v Pamela Fox>every so often you get kind of like a life coach in there.

00:35:41.180 --> 00:35:41.540
<v Pamela Fox>Yes.

00:35:41.900 --> 00:35:42.140
<v Pamela Fox>No,

00:35:42.290 --> 00:35:43.360
<v Michael Kennedy>you do want to, yeah,

00:35:43.880 --> 00:35:45.360
<v Michael Kennedy>like a recruiter.

00:35:45.930 --> 00:35:46.400
<v Michael Kennedy>A recruiter.

00:35:46.560 --> 00:35:46.840
<v Michael Kennedy>Yes,

00:35:47.140 --> 00:35:47.400
<v Pamela Fox>exactly.

00:35:47.740 --> 00:35:49.320
<v Michael Kennedy>A book publisher,

00:35:49.580 --> 00:35:51.060
<v Michael Kennedy>like, hey, would you like to write a book on this?

00:35:51.120 --> 00:35:53.340
<v Michael Kennedy>No, if I want to write a book on it, we've already written a book on it.

00:35:54.340 --> 00:35:57.380
<v Pamela Fox>Yeah, I have this fantasy that one day I'll write a book,

00:35:57.720 --> 00:35:58.340
<v Pamela Fox>so I accept those.

00:35:59.339 --> 00:36:01.400
<v Pamela Fox>But yeah, so I wanted to automate that.

00:36:01.640 --> 00:36:06.480
<v Pamela Fox>So my first experiment was doing a pedantic AI agent with the player and MCP server.

00:36:07.220 --> 00:36:12.360
<v Pamela Fox>And it was like incredibly inefficient because it would like go to the LinkedIn page.

00:36:12.450 --> 00:36:13.480
<v Pamela Fox>It would kind of click around,

00:36:13.720 --> 00:36:14.740
<v Pamela Fox>try to figure out what to do.

00:36:14.940 --> 00:36:18.720
<v Pamela Fox>Like it was like continually reading lots and lots of tokens because it wasn't sure exactly what to do.

00:36:19.220 --> 00:36:20.700
<v Pamela Fox>And it did like eventually succeed.

00:36:21.799 --> 00:36:25.680
<v Pamela Fox>But it wasn't a good experience because one, I had like a limited token budget at that time.

00:36:26.290 --> 00:36:29.180
<v Pamela Fox>And two, it was just like taking a really long time and it was sometimes working,

00:36:29.350 --> 00:36:29.760
<v Pamela Fox>sometimes not.

00:36:29.790 --> 00:36:30.440
<v Pamela Fox>And I was like, well,

00:36:30.820 --> 00:36:31.160
<v Pamela Fox>you know what?

00:36:31.460 --> 00:36:34.040
<v Pamela Fox>I'm trying to do the tasks, the same task every single time,

00:36:34.210 --> 00:36:34.900
<v Pamela Fox>every single day.

00:36:35.200 --> 00:36:38.300
<v Pamela Fox>why don't I just script this using the Python Playwright SDK,

00:36:38.880 --> 00:36:39.020
<v Pamela Fox>right?

00:36:39.440 --> 00:36:39.980
<v Pamela Fox>So I just,

00:36:40.260 --> 00:36:40.740
<v Pamela Fox>you know, I then,

00:36:41.080 --> 00:36:42.880
<v Pamela Fox>you know, scripted it using SDK.

00:36:43.900 --> 00:36:45.460
<v Pamela Fox>You know, at that point, I hand wrote it myself.

00:36:45.780 --> 00:36:48.460
<v Pamela Fox>These days, I just have Copilot open up the browser and figure it out.

00:36:49.260 --> 00:36:49.600
<v Pamela Fox>But then,

00:36:49.920 --> 00:36:51.620
<v Pamela Fox>so then it becomes like much more deterministic.

00:36:51.710 --> 00:36:54.080
<v Pamela Fox>And I only brought the LLM in to make the decision,

00:36:54.340 --> 00:36:54.420
<v Pamela Fox>right?

00:36:54.450 --> 00:36:58.360
<v Pamela Fox>So I have like the Python Playwright SDK code is like opening my LinkedIn,

00:36:59.120 --> 00:37:02.220
<v Pamela Fox>restoring my auth cookies so that I'm like logged in and that had this browser.

00:37:02.630 --> 00:37:03.160
<v Pamela Fox>And then,

00:37:03.320 --> 00:37:03.900
<v Pamela Fox>you know, logging in,

00:37:04.160 --> 00:37:04.780
<v Pamela Fox>checking the requests.

00:37:04.860 --> 00:37:09.700
<v Pamela Fox>it gets a request. Okay. Then that's when it sends the information to an LLM only then. And it uses

00:37:09.940 --> 00:37:11.640
<v Pamela Fox>structured outputs and says, Hey, LLM,

00:37:11.960 --> 00:37:13.700
<v Pamela Fox>send me back some Jason that says,

00:37:13.940 --> 00:37:14.160
<v Pamela Fox>you know,

00:37:14.320 --> 00:37:14.680
<v Pamela Fox>accept,

00:37:14.940 --> 00:37:15.040
<v Pamela Fox>ignore,

00:37:15.300 --> 00:37:19.520
<v Pamela Fox>reject or undecided and give me the reason for it. Right. And I gave it a prompt that said,

00:37:19.660 --> 00:37:20.420
<v Pamela Fox>if it's a recruiter,

00:37:20.880 --> 00:37:23.160
<v Pamela Fox>a life coach or a financial consultant,

00:37:23.700 --> 00:37:23.800
<v Pamela Fox>uh,

00:37:24.220 --> 00:37:24.340
<v Pamela Fox>uh,

00:37:24.540 --> 00:37:25.540
<v Pamela Fox>that's basically the ignore,

00:37:26.020 --> 00:37:26.420
<v Pamela Fox>ignore.

00:37:26.700 --> 00:37:26.780
<v Pamela Fox>Yeah.

00:37:27.380 --> 00:37:27.500
<v Pamela Fox>Um,

00:37:27.940 --> 00:37:28.720
<v Pamela Fox>you know, it's very effective.

00:37:29.280 --> 00:37:29.340
<v Pamela Fox>So,

00:37:29.920 --> 00:37:30.580
<v Pamela Fox>right. So that was,

00:37:30.740 --> 00:37:31.100
<v Pamela Fox>you know, that's,

00:37:31.280 --> 00:37:31.620
<v Pamela Fox>that's,

00:37:31.700 --> 00:37:38.360
<v Pamela Fox>you know that's something considered right like the re like mcp enables these uh you know you to

00:37:38.660 --> 00:37:46.200
<v Pamela Fox>create new mashups new permutations combinations new workflows um that you know you didn't already

00:37:46.340 --> 00:37:51.760
<v Pamela Fox>have in your head so it makes a lot of sense if you are deploying any sort of agent that has more

00:37:51.880 --> 00:37:57.440
<v Pamela Fox>open-ended tasks or that like a user is ultimately going to be used in order to do their uh you know

00:37:57.320 --> 00:37:58.660
<v Pamela Fox>their end tasks, right?

00:37:58.750 --> 00:38:00.760
<v Pamela Fox>So if there's any sort of open-endedness

00:38:01.420 --> 00:38:02.600
<v Pamela Fox>where it may have to do iteration

00:38:02.850 --> 00:38:03.920
<v Pamela Fox>and or shouldn't decide,

00:38:04.070 --> 00:38:04.240
<v Pamela Fox>you know,

00:38:04.270 --> 00:38:05.020
<v Pamela Fox>what it's going to do next,

00:38:05.680 --> 00:38:07.260
<v Pamela Fox>then an MSP server makes sense.

00:38:07.680 --> 00:38:09.240
<v Pamela Fox>But if you're doing the same thing

00:38:09.620 --> 00:38:10.400
<v Pamela Fox>every single time,

00:38:10.730 --> 00:38:11.520
<v Pamela Fox>and there's, you know,

00:38:11.630 --> 00:38:12.040
<v Pamela Fox>even like,

00:38:12.260 --> 00:38:12.420
<v Pamela Fox>you know,

00:38:12.560 --> 00:38:13.360
<v Pamela Fox>in the case of mine,

00:38:13.360 --> 00:38:14.240
<v Pamela Fox>where it's like all you're doing

00:38:14.270 --> 00:38:15.500
<v Pamela Fox>is tracing network requests,

00:38:16.000 --> 00:38:17.360
<v Pamela Fox>it's just not necessary

00:38:17.720 --> 00:38:18.440
<v Pamela Fox>to use MSP server

00:38:18.610 --> 00:38:20.680
<v Pamela Fox>and it can add more non-determinism

00:38:21.080 --> 00:38:21.740
<v Pamela Fox>to a workflow.

00:38:22.140 --> 00:38:22.860
<v Michael Kennedy>Yeah, and if you're worried

00:38:22.980 --> 00:38:23.780
<v Michael Kennedy>about it going crazy

00:38:24.020 --> 00:38:25.700
<v Michael Kennedy>and just instead of connecting,

00:38:25.920 --> 00:38:26.120
<v Michael Kennedy>Oh,

00:38:26.600 --> 00:38:27.620
<v Michael Kennedy>she wants to connect with people.

00:38:27.800 --> 00:38:30.220
<v Michael Kennedy>Let me find everyone on LinkedIn and add them as a connection.

00:38:30.440 --> 00:38:30.700
<v Michael Kennedy>You know what I mean?

00:38:30.800 --> 00:38:31.660
<v Michael Kennedy>Like something like that.

00:38:31.870 --> 00:38:33.300
<v Michael Kennedy>You could think, oh, the AI went crazy.

00:38:33.680 --> 00:38:37.700
<v Michael Kennedy>What you've given it is a much more constrained and deterministic deal, right?

00:38:37.960 --> 00:38:39.340
<v Michael Kennedy>Like it can only look at yes,

00:38:39.500 --> 00:38:39.620
<v Michael Kennedy>no.

00:38:40.100 --> 00:38:40.280
<v Michael Kennedy>I mean,

00:38:40.640 --> 00:38:42.520
<v Michael Kennedy>today almost would be a Jev sort of thing.

00:38:42.780 --> 00:38:45.100
<v Pamela Fox>Yeah, I could totally sub in a Jev model point,

00:38:45.260 --> 00:38:45.360
<v Pamela Fox>right?

00:38:45.500 --> 00:38:48.100
<v Pamela Fox>Like it would be a good scenario for a Jev model.

00:38:48.280 --> 00:38:48.500
<v Pamela Fox>Generally,

00:38:48.740 --> 00:38:49.640
<v Pamela Fox>like that's a whole other discussion.

00:38:49.750 --> 00:38:50.080
<v Pamela Fox>But yeah,

00:38:50.260 --> 00:38:53.700
<v Pamela Fox>generally, if anywhere we're using structured outputs where the result is a Boolean

00:38:53.720 --> 00:38:54.220
<v Pamela Fox>or a choice,

00:38:54.620 --> 00:38:55.960
<v Pamela Fox>you can go ahead and sub in

00:38:56.260 --> 00:38:56.740
<v Pamela Fox>JAP model,

00:38:56.980 --> 00:38:57.560
<v Pamela Fox>decisions API,

00:38:58.380 --> 00:38:59.880
<v Pamela Fox>Microsoft, we just launched decision one,

00:39:00.960 --> 00:39:02.420
<v Pamela Fox>you know, a decision model.

00:39:02.820 --> 00:39:05.380
<v Pamela Fox>And yeah, so then you don't even need a full LLM.

00:39:05.500 --> 00:39:05.760
<v Michael Kennedy>Yeah,

00:39:06.020 --> 00:39:06.140
<v Michael Kennedy>exactly.

00:39:06.620 --> 00:39:08.640
<v Michael Kennedy>But I do think a lot of times just giving an LLM

00:39:08.780 --> 00:39:10.380
<v Michael Kennedy>a chance to like, I got it close,

00:39:10.840 --> 00:39:11.740
<v Michael Kennedy>but let's go there.

00:39:11.860 --> 00:39:14.240
<v Michael Kennedy>I'll give you an example from my world

00:39:14.360 --> 00:39:15.540
<v Michael Kennedy>that I found really helpful

00:39:16.020 --> 00:39:18.640
<v Michael Kennedy>is I have Hermes chilling on my Mac mini

00:39:18.840 --> 00:39:20.160
<v Michael Kennedy>just doing a bunch of stuff

00:39:20.600 --> 00:39:22.680
<v Michael Kennedy>and two YouTube channels

00:39:22.700 --> 00:39:25.520
<v Michael Kennedy>and they really are just sort of replications of the podcast.

00:39:25.880 --> 00:39:27.360
<v Michael Kennedy>So like the description is,

00:39:27.700 --> 00:39:28.460
<v Michael Kennedy>it's a structured thing,

00:39:28.540 --> 00:39:30.860
<v Michael Kennedy>but it comes from a whole bunch of different pieces of data

00:39:31.020 --> 00:39:32.240
<v Michael Kennedy>out of the podcast and so on.

00:39:32.580 --> 00:39:34.360
<v Michael Kennedy>But especially when I was live streaming more,

00:39:34.740 --> 00:39:36.380
<v Michael Kennedy>it would be the stuff will go up.

00:39:36.560 --> 00:39:38.360
<v Michael Kennedy>And then eventually when the results,

00:39:38.680 --> 00:39:40.240
<v Michael Kennedy>like all the details of the episode are together,

00:39:40.280 --> 00:39:41.920
<v Michael Kennedy>I got to go find all of that, the transcript,

00:39:42.240 --> 00:39:42.700
<v Michael Kennedy>the description,

00:39:42.880 --> 00:39:43.460
<v Michael Kennedy>the guest, the links,

00:39:43.560 --> 00:39:44.680
<v Michael Kennedy>and then go and put them back.

00:39:45.060 --> 00:39:46.060
<v Michael Kennedy>And I just sort of,

00:39:46.540 --> 00:39:47.520
<v Michael Kennedy>I don't know, I lost track of that.

00:39:47.760 --> 00:39:50.900
<v Michael Kennedy>And I had like two or 300 videos on each channel

00:39:50.920 --> 00:39:52.500
<v Michael Kennedy>that were just episode seven,

00:39:53.100 --> 00:39:53.800
<v Michael Kennedy>episode 200.

00:39:53.960 --> 00:39:54.260
<v Michael Kennedy>You know what I mean?

00:39:54.360 --> 00:39:54.880
<v Michael Kennedy>Like, oh no,

00:39:55.240 --> 00:39:57.700
<v Michael Kennedy>I can't, I cannot navigate the YouTube studio.

00:39:57.980 --> 00:40:00.120
<v Michael Kennedy>It's so slow on every little step.

00:40:00.140 --> 00:40:02.880
<v Michael Kennedy>I can't go and do like 400 edits of this.

00:40:03.140 --> 00:40:05.060
<v Michael Kennedy>And so I wanted my Hermes to just,

00:40:05.420 --> 00:40:07.360
<v Michael Kennedy>anytime that sees something appear in the RSS feed

00:40:07.360 --> 00:40:10.060
<v Michael Kennedy>for the podcast to instantly use the MCP server,

00:40:10.240 --> 00:40:12.960
<v Michael Kennedy>find the details and rewrite the YouTube channel.

00:40:13.380 --> 00:40:15.640
<v Michael Kennedy>Well, I don't want to just give it full access to my YouTube.

00:40:15.820 --> 00:40:19.859
<v Michael Kennedy>So what I did is I used Claude Code to build a Python library

00:40:19.860 --> 00:40:22.600
<v Michael Kennedy>that does the specific things I want it to do to YouTube.

00:40:23.080 --> 00:40:24.560
<v Michael Kennedy>And I just gave Hermes the wheel.

00:40:24.620 --> 00:40:25.580
<v Michael Kennedy>I said, here's the wheel.

00:40:25.680 --> 00:40:27.140
<v Michael Kennedy>You install this if you need to.

00:40:27.440 --> 00:40:29.060
<v Michael Kennedy>Then you manage the YouTube channel through it.

00:40:29.320 --> 00:40:30.020
<v Michael Kennedy>And it does.

00:40:30.160 --> 00:40:33.000
<v Michael Kennedy>And it's gone through all my hundreds of videos

00:40:33.100 --> 00:40:33.760
<v Michael Kennedy>are all up to date.

00:40:33.880 --> 00:40:34.140
<v Michael Kennedy>It's perfect.

00:40:34.480 --> 00:40:35.420
<v Michael Kennedy>And if I change it on the website,

00:40:35.700 --> 00:40:37.940
<v Michael Kennedy>it automatically changes on YouTube like 15 minutes later.

00:40:38.180 --> 00:40:38.440
<v Michael Kennedy>Crazy.

00:40:38.780 --> 00:40:40.580
<v Pamela Fox>I think that's something that we can take advantage

00:40:40.860 --> 00:40:43.760
<v Pamela Fox>as Python developers is that it's easy for us

00:40:43.760 --> 00:40:46.620
<v Pamela Fox>to make our own tools that have a more restricted surface.

00:40:46.800 --> 00:40:47.240
<v Pamela Fox>For example,

00:40:47.480 --> 00:40:49.840
<v Pamela Fox>I keep meaning to VibeCode

00:40:49.840 --> 00:40:52.080
<v Pamela Fox>MSP server for my Pilates booking website,

00:40:52.600 --> 00:40:52.720
<v Pamela Fox>right?

00:40:52.920 --> 00:40:55.480
<v Pamela Fox>And I would just expose only the tools that I want, right?

00:40:55.500 --> 00:40:58.740
<v Pamela Fox>Like I wouldn't have, you know, I wouldn't have like my account.

00:40:59.120 --> 00:41:03.120
<v Pamela Fox>I would just have a, you know, like list this week sessions and book a session,

00:41:03.380 --> 00:41:03.480
<v Pamela Fox>right?

00:41:03.600 --> 00:41:03.820
<v Pamela Fox>So,

00:41:04.240 --> 00:41:04.600
<v Pamela Fox>you know, for,

00:41:05.180 --> 00:41:06.900
<v Pamela Fox>you know, especially for us as Python developers,

00:41:07.120 --> 00:41:07.680
<v Pamela Fox>this should be really

00:41:07.820 --> 00:41:07.920
<v Pamela Fox>easy.

00:41:08.000 --> 00:41:08.660
<v Pamela Fox>We just,

00:41:09.020 --> 00:41:10.020
<v Pamela Fox>you know, prompt and be like, hey,

00:41:10.360 --> 00:41:12.480
<v Pamela Fox>you know, make this MSP server that does this

00:41:12.620 --> 00:41:12.940
<v Pamela Fox>and this,

00:41:13.980 --> 00:41:15.760
<v Pamela Fox>you know, and the hardest part is like auth,

00:41:15.920 --> 00:41:16.260
<v Pamela Fox>but there's,

00:41:16.420 --> 00:41:16.740
<v Pamela Fox>you know, there's

00:41:16.740 --> 00:41:17.940
<v Pamela Fox>ways to set that up.

00:41:18.140 --> 00:41:22.500
<v Pamela Fox>And then we can like feel confident pointing our agents at these tools,

00:41:22.880 --> 00:41:24.740
<v Pamela Fox>these servers or CLIs or however you've done them,

00:41:25.200 --> 00:41:30.620
<v Pamela Fox>and not worry about something bad happening,

00:41:30.820 --> 00:41:30.920
<v Pamela Fox>right?

00:41:31.160 --> 00:41:36.900
<v Pamela Fox>I know like many people are just kind of YOLOing with the current rage of Muse and Grok Pot and that.

00:41:36.980 --> 00:41:38.700
<v Pamela Fox>But then, of course, every day we see this like,

00:41:38.900 --> 00:41:41.380
<v Pamela Fox>you know, horrible thing that happened to someone because of that.

00:41:41.780 --> 00:41:43.840
<v Michael Kennedy>It RMRF'd our production database.

00:41:44.100 --> 00:41:46.460
<v Michael Kennedy>Why did you give it access to your production database?

00:41:46.600 --> 00:41:47.060
<v Michael Kennedy>What are you doing?

00:41:48.000 --> 00:41:48.220
<v Pamela Fox>So,

00:41:48.500 --> 00:41:52.060
<v Pamela Fox>you know, we we're still in an advantage as developers where,

00:41:52.640 --> 00:41:58.240
<v Pamela Fox>you know, if we don't have already access to tools that allow us to do restriction,

00:41:58.780 --> 00:41:59.380
<v Pamela Fox>then we can,

00:41:59.740 --> 00:42:02.180
<v Pamela Fox>you know, we can buy our own tools for that.

00:42:02.300 --> 00:42:04.360
<v Pamela Fox>Many of these tools, like, you know, with MCP servers,

00:42:04.560 --> 00:42:09.660
<v Pamela Fox>a lot of them will have a bunch of tools and you could decide to give your agent access to just some of them and not others.

00:42:09.920 --> 00:42:10.000
<v Pamela Fox>Right.

00:42:10.160 --> 00:42:10.220
<v Pamela Fox>So,

00:42:10.500 --> 00:42:10.620
<v Pamela Fox>you know,

00:42:10.960 --> 00:42:14.640
<v Pamela Fox>MCP servers will actually like annotate which of the tools are read only.

00:42:14.800 --> 00:42:15.520
<v Pamela Fox>And you might say, hey,

00:42:15.980 --> 00:42:18.180
<v Pamela Fox>you know, you can only access the read only tools.

00:42:19.100 --> 00:42:22.020
<v Pamela Fox>And with like the GitHub MCP server, it has this notion of tool sets.

00:42:22.160 --> 00:42:23.240
<v Pamela Fox>And you can say you can say like, hey,

00:42:23.780 --> 00:42:25.140
<v Pamela Fox>you can you're only allowed to,

00:42:25.340 --> 00:42:25.460
<v Pamela Fox>you know,

00:42:25.620 --> 00:42:26.620
<v Pamela Fox>use such and such tool set.

00:42:26.920 --> 00:42:29.100
<v Michael Kennedy>A little bit like restricting the job that you give them.

00:42:29.560 --> 00:42:30.700
<v Michael Kennedy>Because when they come back,

00:42:30.880 --> 00:42:32.920
<v Michael Kennedy>it goes to the MCP and it says, what can I do?

00:42:32.960 --> 00:42:34.680
<v Michael Kennedy>And it's like, oh, well, here's 200 things you can do.

00:42:35.400 --> 00:42:38.360
<v Michael Kennedy>You don't necessarily want it to do those 200 things just randomly.

00:42:38.440 --> 00:42:40.720
<v Michael Kennedy>It might be like, like you said, list schedules,

00:42:41.200 --> 00:42:41.920
<v Michael Kennedy>check my free time,

00:42:42.220 --> 00:42:42.860
<v Michael Kennedy>book me this thing.

00:42:43.100 --> 00:42:44.620
<v Michael Kennedy>You get three or four tools,

00:42:44.860 --> 00:42:45.400
<v Michael Kennedy>you know what I mean?

00:42:45.520 --> 00:42:46.020
<v Michael Kennedy>And so,

00:42:46.280 --> 00:42:49.040
<v Michael Kennedy>yeah, you have a talk on just stricter schemas,

00:42:49.220 --> 00:42:49.680
<v Michael Kennedy>increased reliability,

00:42:49.980 --> 00:42:53.100
<v Michael Kennedy>and then also on directing them to the, like, oh, the filter of the tools one.

00:42:53.180 --> 00:42:53.380
<v Michael Kennedy>That's it.

00:42:53.500 --> 00:42:54.980
<v Michael Kennedy>Filtering tools from MCP servers.

00:42:56.060 --> 00:42:57.780
<v Michael Kennedy>Basically how you can save tokens,

00:42:57.970 --> 00:42:58.460
<v Michael Kennedy>make it safer.

00:42:59.120 --> 00:43:01.060
<v Michael Kennedy>There's some tools in Copilot,

00:43:01.360 --> 00:43:05.940
<v Michael Kennedy>and this might also apply if you run Codex or Claude

00:43:06.260 --> 00:43:07.600
<v Michael Kennedy>in the chat window in VS Code.

00:43:07.670 --> 00:43:08.140
<v Michael Kennedy>I can't remember.

00:43:08.460 --> 00:43:08.680
<v Pamela Fox>Yeah.

00:43:08.790 --> 00:43:12.260
<v Pamela Fox>So when I wrote this blog post here about filtering tools,

00:43:12.540 --> 00:43:13.140
<v Pamela Fox>this was,

00:43:13.340 --> 00:43:13.740
<v Pamela Fox>what, five?

00:43:15.320 --> 00:43:15.500
<v Pamela Fox>Ancient.

00:43:15.860 --> 00:43:16.140
<v Michael Kennedy>It's ancient.

00:43:16.250 --> 00:43:16.360
<v Pamela Fox>Ancient.

00:43:16.760 --> 00:43:17.520
<v Pamela Fox>So at that time,

00:43:17.760 --> 00:43:18.560
<v Pamela Fox>like in early days,

00:43:18.780 --> 00:43:19.420
<v Pamela Fox>VS Code,

00:43:19.570 --> 00:43:22.580
<v Pamela Fox>actually if you selected more than I think it was like 175 tools,

00:43:22.780 --> 00:43:24.780
<v Pamela Fox>you'd like get this warning that's like, you have more than,

00:43:25.300 --> 00:43:25.520
<v Pamela Fox>you know,

00:43:26.090 --> 00:43:27.180
<v Pamela Fox>this many tools selected,

00:43:27.400 --> 00:43:30.760
<v Pamela Fox>like you may want to reduce them, right? And everyone would get so stressed out, like trying

00:43:30.860 --> 00:43:33.600
<v Pamela Fox>to decide which of them to like turn off and turn on.

00:43:34.040 --> 00:43:36.380
<v Pamela Fox>Now, a few things have changed since that time.

00:43:37.020 --> 00:43:38.420
<v Pamela Fox>One, obviously, our models have gotten,

00:43:38.580 --> 00:43:39.580
<v Pamela Fox>but two,

00:43:39.830 --> 00:43:42.780
<v Pamela Fox>the models have tool search and deferred tool

00:43:42.970 --> 00:43:48.820
<v Pamela Fox>loading built in. So both the new OpenAI models and Anthropic models, like from the last six months

00:43:48.840 --> 00:43:49.160
<v Pamela Fox>or so,

00:43:49.320 --> 00:43:54.340
<v Pamela Fox>they have a built-in capability that's actually was like trained into them as part

00:43:54.340 --> 00:43:56.600
<v Pamela Fox>of their RLHF, their post-training,

00:43:56.800 --> 00:44:01.320
<v Pamela Fox>where they know how to search for a tool and to

00:44:01.520 --> 00:44:02.140
<v Pamela Fox>load in a tool,

00:44:02.420 --> 00:44:07.680
<v Pamela Fox>like to use a deferred tool if they don't see a tool fully loaded in,

00:44:07.870 --> 00:44:08.020
<v Pamela Fox>right?

00:44:08.110 --> 00:44:09.960
<v Pamela Fox>So a lot of the modern agents will,

00:44:10.250 --> 00:44:12.280
<v Pamela Fox>if they're seeing that you're using a model,

00:44:12.620 --> 00:44:12.740
<v Pamela Fox>like our

00:44:12.920 --> 00:44:13.400
<v Pamela Fox>OpenAI model,

00:44:13.910 --> 00:44:17.400
<v Pamela Fox>they may choose to use that deferred tool.

00:44:17.780 --> 00:44:18.420
<v Pamela Fox>So these days,

00:44:18.840 --> 00:44:25.580
<v Pamela Fox>I don't really worry as much about having too many tools turned on because I know that there's,

00:44:25.740 --> 00:44:27.840
<v Pamela Fox>you know, much better handling of many tools.

00:44:29.120 --> 00:44:29.240
<v Pamela Fox>However,

00:44:29.640 --> 00:44:29.880
<v Pamela Fox>you know,

00:44:30.480 --> 00:44:34.000
<v Pamela Fox>if you like change your model to like a much worse model or something like, you know, then,

00:44:35.160 --> 00:44:36.480
<v Pamela Fox>you know, then you'll differ.

00:44:37.340 --> 00:44:39.680
<v Michael Kennedy>Yeah, I played around with one of the smaller,

00:44:40.100 --> 00:44:41.320
<v Michael Kennedy>more self-hosted like models.

00:44:41.380 --> 00:44:43.700
<v Michael Kennedy>It wasn't quite that, but it wasn't the frontier ones.

00:44:44.020 --> 00:44:45.660
<v Michael Kennedy>And I gave it basically this project.

00:44:45.780 --> 00:44:47.140
<v Michael Kennedy>I want you to change part of this.

00:44:47.280 --> 00:44:48.480
<v Michael Kennedy>It's like a little bit later, it's like,

00:44:48.660 --> 00:44:50.260
<v Michael Kennedy>I've removed all of this from your app.

00:44:50.280 --> 00:44:50.900
<v Michael Kennedy>I'm like, no, no, no.

00:44:51.040 --> 00:44:52.460
<v Michael Kennedy>Why did you think you'd delete all?

00:44:52.540 --> 00:44:54.600
<v Michael Kennedy>I just wanted you to take a little part of it.

00:44:55.900 --> 00:44:59.000
<v Michael Kennedy>And it just made me remember like, oh, yeah, it used to be like that all the time.

00:44:59.140 --> 00:45:00.840
<v Michael Kennedy>It would just go crazy and do weird stuff.

00:45:01.080 --> 00:45:04.680
<v Michael Kennedy>And that doesn't really happen very much with the top tier models these days.

00:45:05.700 --> 00:45:07.420
<v Michael Kennedy>You're a little bit going to let it figure it out.

00:45:07.560 --> 00:45:11.180
<v Michael Kennedy>I think that's some of the newer guidance is it's smart enough that you're just going to be in its way.

00:45:12.180 --> 00:45:16.480
<v Pamela Fox>Yeah, it means like when you're designing MSP servers and skills,

00:45:16.640 --> 00:45:17.260
<v Pamela Fox>especially skills,

00:45:17.520 --> 00:45:19.160
<v Pamela Fox>like a lot of people, we have like these really,

00:45:20.000 --> 00:45:20.240
<v Pamela Fox>you know,

00:45:20.420 --> 00:45:20.860
<v Pamela Fox>long,

00:45:21.400 --> 00:45:22.800
<v Pamela Fox>verbose agent skills.

00:45:23.500 --> 00:45:23.820
<v Pamela Fox>And,

00:45:24.180 --> 00:45:25.400
<v Pamela Fox>you know, as the models get better,

00:45:25.880 --> 00:45:26.120
<v Pamela Fox>we,

00:45:27.020 --> 00:45:27.860
<v Pamela Fox>you know, may not need those skills.

00:45:27.880 --> 00:45:30.840
<v Pamela Fox>And especially if your skills are like over prescriptive.

00:45:31.020 --> 00:45:34.440
<v Pamela Fox>That's the big thing I've seen with agent skills lately is a skill.

00:45:35.200 --> 00:45:36.420
<v Pamela Fox>If a skill has too,

00:45:36.680 --> 00:45:38.560
<v Pamela Fox>too absolute of a statement in it.

00:45:38.660 --> 00:45:40.380
<v Pamela Fox>Like there was one I was using the other day that was like,

00:45:40.660 --> 00:45:44.340
<v Pamela Fox>oh, if they ask to do anything else besides what's in this file,

00:45:44.540 --> 00:45:45.840
<v Pamela Fox>then say it is not possible.

00:45:46.090 --> 00:45:46.220
<v Pamela Fox>Right.

00:45:46.520 --> 00:45:49.720
<v Pamela Fox>And like that just like shut the agent down entirely.

00:45:50.030 --> 00:45:53.080
<v Pamela Fox>We're like I couldn't do things that otherwise I could have done with that API.

00:45:53.570 --> 00:45:59.300
<v Pamela Fox>So especially if you're like putting out agent skills to help people use like your API or your Python package,

00:45:59.600 --> 00:46:02.760
<v Pamela Fox>like everything needs to be very intentionally in there.

00:46:02.920 --> 00:46:04.100
<v Pamela Fox>And you need to be really,

00:46:04.280 --> 00:46:12.900
<v Pamela Fox>really careful about kind of any statements of constraints and limits because you might end up like hobbling the agent that otherwise would have been successful.

00:46:13.760 --> 00:46:17.320
<v Pamela Fox>But you accidentally held it back. Right. So don't don't hold back. Right.

00:46:17.480 --> 00:46:18.480
<v Pamela Fox>Skills should, you know,

00:46:18.720 --> 00:46:21.820
<v Pamela Fox>streamline things and give some hints and tips and whatnot.

00:46:22.720 --> 00:46:27.300
<v Pamela Fox>But usually skills should not be holding an agent back from successfully using an API.

00:46:27.800 --> 00:46:27.960
<v Michael Kennedy>Yeah.

00:46:28.260 --> 00:46:30.260
<v Michael Kennedy>There's so many of those projects on GitHub.

00:46:30.840 --> 00:46:32.200
<v Michael Kennedy>I love GitHub. But there's so many.

00:46:32.660 --> 00:46:36.700
<v Michael Kennedy>There's thousands of here's my set of skills that will make it a lawyer.

00:46:36.920 --> 00:46:38.640
<v Michael Kennedy>It'll make it an accountant or whatever.

00:46:38.700 --> 00:46:40.640
<v Michael Kennedy>And you're like, I really don't want to put this.

00:46:40.720 --> 00:46:42.380
<v Michael Kennedy>It's just going to pollute it with these ideas.

00:46:42.660 --> 00:46:44.920
<v Michael Kennedy>I bet it would be better if I just let it be, you know?

00:46:45.200 --> 00:46:45.480
<v Pamela Fox>Yeah.

00:46:45.920 --> 00:46:48.320
<v Pamela Fox>I mean, I'm happy to use my own skills.

00:46:48.460 --> 00:46:53.840
<v Pamela Fox>I think what I encourage people is find skills that people are talking about that you're in shit in.

00:46:54.040 --> 00:46:56.420
<v Pamela Fox>Fork them and then make them your own, right?

00:46:56.680 --> 00:47:02.700
<v Pamela Fox>Because like there's very few skills that I use from other people because,

00:47:03.100 --> 00:47:03.220
<v Pamela Fox>you know,

00:47:03.440 --> 00:47:04.180
<v Pamela Fox>I'm very opinionated.

00:47:04.280 --> 00:47:06.040
<v Pamela Fox>I want my workflows to work great for me.

00:47:06.240 --> 00:47:09.020
<v Pamela Fox>So I'm very happy with the skills that I maintain.

00:47:09.220 --> 00:47:13.320
<v Pamela Fox>Right. So I have presentation skills is my biggest repo of skills.

00:47:13.380 --> 00:47:16.020
<v Pamela Fox>And it's all just things related to making presentations like,

00:47:16.980 --> 00:47:19.680
<v Pamela Fox>you know, generating slide decks and images for them,

00:47:19.880 --> 00:47:21.620
<v Pamela Fox>making transcripts, lots of stuff with YouTube.

00:47:21.940 --> 00:47:22.040
<v Pamela Fox>Right.

00:47:22.240 --> 00:47:24.000
<v Pamela Fox>Like transcribing them after.

00:47:24.520 --> 00:47:29.060
<v Pamela Fox>And then I have some other skills that have to do with like Azure deployment and stuff.

00:47:29.440 --> 00:47:32.060
<v Pamela Fox>So I'm happy with those skills because I have a lot of control over them.

00:47:32.070 --> 00:47:37.620
<v Pamela Fox>But it's so hard to use someone else's skill and have it be a great experience for you.

00:47:37.800 --> 00:47:39.800
<v Pamela Fox>So find a skill you like and then fork it.

00:47:40.000 --> 00:47:40.700
<v Pamela Fox>Make it your own.

00:47:40.980 --> 00:47:41.160
<v Michael Kennedy>Yeah,

00:47:41.360 --> 00:47:41.800
<v Michael Kennedy>that's great advice.

00:47:42.280 --> 00:47:42.460
<v Michael Kennedy>Absolutely.

00:47:42.880 --> 00:47:43.100
<v Michael Kennedy>Now,

00:47:43.440 --> 00:47:45.560
<v Michael Kennedy>let's talk CLIs for a minute.

00:47:45.890 --> 00:47:49.680
<v Michael Kennedy>I think having CLIs used to be something that was just cool.

00:47:49.880 --> 00:47:52.480
<v Michael Kennedy>So you could maybe script it or you could automate it.

00:47:52.580 --> 00:47:54.780
<v Michael Kennedy>Like, I want to be able to do this thing with something.

00:47:55.360 --> 00:47:56.900
<v Michael Kennedy>And if I make a CLI,

00:47:57.200 --> 00:48:00.560
<v Michael Kennedy>then my Bash or my Ansible script or whatever can run it.

00:48:00.820 --> 00:48:02.120
<v Michael Kennedy>So that's why we're going to have a CLI.

00:48:02.180 --> 00:48:02.820
<v Michael Kennedy>And I think, honestly,

00:48:03.160 --> 00:48:08.680
<v Michael Kennedy>CLIs these days are a really interesting escape hatch for agents and things like that.

00:48:08.700 --> 00:48:09.640
<v Michael Kennedy>They're like, for some reason,

00:48:09.840 --> 00:48:12.160
<v Michael Kennedy>they don't connect or they can't connect to an MCP.

00:48:12.400 --> 00:48:17.380
<v Michael Kennedy>Like, maybe it's an app that only would have an MCP when it's running, but you only run it for a moment.

00:48:17.680 --> 00:48:20.720
<v Michael Kennedy>So maybe give us your thoughts on this CLI story.

00:48:21.240 --> 00:48:24.700
<v Pamela Fox>Yes, we have to have the required MCP versus CLI debate.

00:48:25.860 --> 00:48:26.140
<v Michael Kennedy>We do.

00:48:26.240 --> 00:48:27.700
<v Michael Kennedy>It wouldn't be complete if we didn't do it.

00:48:27.900 --> 00:48:29.100
<v Pamela Fox>It's contractually obliged.

00:48:29.920 --> 00:48:30.400
<v Pamela Fox>So,

00:48:31.400 --> 00:48:32.240
<v Pamela Fox>yeah, it's very interesting,

00:48:32.440 --> 00:48:32.560
<v Pamela Fox>actually.

00:48:32.840 --> 00:48:32.920
<v Pamela Fox>Like,

00:48:33.740 --> 00:48:37.400
<v Pamela Fox>my colleague has a project that's now called Scope.

00:48:37.440 --> 00:48:38.300
<v Pamela Fox>And in this project,

00:48:38.520 --> 00:48:44.700
<v Pamela Fox>they actually will set up scenarios to see how well an agent succeeds at a particular task in a particular environment.

00:48:45.020 --> 00:48:50.060
<v Pamela Fox>And the example that he gave in a talk the other day was using the GitHub CLI,

00:48:50.440 --> 00:48:51.120
<v Pamela Fox>GH command.

00:48:51.500 --> 00:48:57.020
<v Pamela Fox>and the GitHub MCP server and then just nothing at all compared to see like how long would it take?

00:48:57.130 --> 00:48:58.940
<v Pamela Fox>How successful would it be? Like would it?

00:48:59.670 --> 00:49:02.120
<v Pamela Fox>So in this case, it was actually fixing a bug on one of my repos,

00:49:03.580 --> 00:49:04.620
<v Pamela Fox>which I didn't know he was going to do.

00:49:04.650 --> 00:49:06.880
<v Pamela Fox>So I was surprised to see this PR the morning of his talk.

00:49:07.060 --> 00:49:09.220
<v Pamela Fox>But yeah, so he's,

00:49:09.480 --> 00:49:09.640
<v Pamela Fox>you know,

00:49:09.800 --> 00:49:12.180
<v Pamela Fox>so it's comparing to see, you know, token efficiency,

00:49:12.490 --> 00:49:12.580
<v Pamela Fox>success,

00:49:13.500 --> 00:49:14.880
<v Pamela Fox>number of tool calls, all that stuff.

00:49:15.100 --> 00:49:15.160
<v Pamela Fox>Right.

00:49:15.640 --> 00:49:21.160
<v Pamela Fox>Now, the results from that particular experiment was that the GitHub CLI was the best, right?

00:49:21.420 --> 00:49:23.520
<v Pamela Fox>And I think a lot of people would actually predict that.

00:49:23.680 --> 00:49:29.740
<v Pamela Fox>It required less tokens overall and was successful more of the time.

00:49:30.680 --> 00:49:31.400
<v Pamela Fox>But why is that?

00:49:31.660 --> 00:49:34.680
<v Pamela Fox>Part of that is because the agents, like LLMs,

00:49:34.860 --> 00:49:36.040
<v Pamela Fox>they know the GitHub CLI.

00:49:36.160 --> 00:49:37.620
<v Pamela Fox>The GitHub CLI is deep,

00:49:37.880 --> 00:49:38.780
<v Pamela Fox>deep in the weights,

00:49:39.000 --> 00:49:39.080
<v Pamela Fox>right?

00:49:39.220 --> 00:49:39.980
<v Pamela Fox>So they're very,

00:49:40.160 --> 00:49:42.320
<v Pamela Fox>very good at using the GitHub CLI.

00:49:42.480 --> 00:49:44.600
<v Pamela Fox>they really don't know the GitHub MCP server much at all,

00:49:44.710 --> 00:49:44.820
<v Pamela Fox>especially,

00:49:45.050 --> 00:49:46.120
<v Pamela Fox>you know, it changes over time.

00:49:46.580 --> 00:49:47.380
<v Pamela Fox>So it's very,

00:49:47.580 --> 00:49:49.120
<v Pamela Fox>it's really not surprising

00:49:49.760 --> 00:49:53.760
<v Pamela Fox>that they would do particularly well at using that CLI.

00:49:54.620 --> 00:49:54.840
<v Pamela Fox>So,

00:49:55.140 --> 00:49:57.100
<v Pamela Fox>you know, so the takeaway from that experiment

00:49:57.380 --> 00:49:58.760
<v Pamela Fox>is like in that particular example,

00:49:59.160 --> 00:50:00.320
<v Pamela Fox>yeah, it was more successful with CLI

00:50:00.540 --> 00:50:02.580
<v Pamela Fox>and there were some good kind of just specific learnings

00:50:02.640 --> 00:50:05.100
<v Pamela Fox>for the GitHub MCP server about what it struggled with

00:50:05.110 --> 00:50:06.240
<v Pamela Fox>when it was calling that server.

00:50:07.140 --> 00:50:09.840
<v Pamela Fox>But generally what you should be doing is like ideally

00:50:10.240 --> 00:50:11.400
<v Pamela Fox>like always offering everything,

00:50:11.800 --> 00:50:11.880
<v Pamela Fox>right?

00:50:11.980 --> 00:50:14.240
<v Pamela Fox>So you should have a CLI probably.

00:50:14.440 --> 00:50:15.380
<v Pamela Fox>You should have an API,

00:50:15.560 --> 00:50:16.400
<v Pamela Fox>like a traditional API.

00:50:16.520 --> 00:50:17.840
<v Pamela Fox>Like we're not getting rid of APIs.

00:50:18.020 --> 00:50:19.520
<v Pamela Fox>Like APIs for HTTP services,

00:50:19.860 --> 00:50:21.580
<v Pamela Fox>like you still want an API,

00:50:22.080 --> 00:50:22.560
<v Pamela Fox>an SDK,

00:50:23.080 --> 00:50:23.720
<v Pamela Fox>a CLI.

00:50:24.580 --> 00:50:25.720
<v Pamela Fox>An MCP server,

00:50:26.240 --> 00:50:26.440
<v Pamela Fox>you know,

00:50:26.500 --> 00:50:29.160
<v Pamela Fox>is something that you've ideally optimized for agents.

00:50:29.200 --> 00:50:31.340
<v Pamela Fox>So you should be like running these tests with agents

00:50:31.400 --> 00:50:31.740
<v Pamela Fox>and saying,

00:50:31.900 --> 00:50:31.980
<v Pamela Fox>hey,

00:50:32.180 --> 00:50:32.800
<v Pamela Fox>for this scenario,

00:50:33.220 --> 00:50:34.560
<v Pamela Fox>for all these range of scenarios,

00:50:35.460 --> 00:50:37.960
<v Pamela Fox>how successful are the agents when,

00:50:38.640 --> 00:50:39.880
<v Pamela Fox>you know, when using our MCP server?

00:50:40.080 --> 00:50:41.100
<v Pamela Fox>And also with agent skills,

00:50:41.260 --> 00:50:43.000
<v Pamela Fox>Like we do the same test with agent skills.

00:50:43.880 --> 00:50:44.320
<v Pamela Fox>So these are,

00:50:44.480 --> 00:50:45.520
<v Pamela Fox>you know, think of these as,

00:50:45.700 --> 00:50:47.760
<v Pamela Fox>you know, you're developing tools for agents

00:50:48.320 --> 00:50:50.900
<v Pamela Fox>and you really want to see how successful agents.

00:50:51.040 --> 00:50:52.020
<v Pamela Fox>You can do the same with CLI.

00:50:52.130 --> 00:50:56.260
<v Pamela Fox>Like you could, there are definitely people who are trying to optimize their CLIs more

00:50:56.580 --> 00:50:57.260
<v Pamela Fox>for agents.

00:50:57.350 --> 00:51:02.380
<v Pamela Fox>Like there was a great talk at PyAI last year from the folks at Astral where they were talking

00:51:02.430 --> 00:51:05.600
<v Pamela Fox>about that there was a lot of improvements they were going to make to like rough that

00:51:05.680 --> 00:51:06.800
<v Pamela Fox>made it better for humans.

00:51:07.330 --> 00:51:08.560
<v Pamela Fox>But now they're like, well, what's the point?

00:51:08.860 --> 00:51:10.400
<v Pamela Fox>Humans aren't the ones running rough anymore.

00:51:10.940 --> 00:51:11.760
<v Pamela Fox>agents are running rough.

00:51:12.000 --> 00:51:14.240
<v Pamela Fox>They have all these ideas now about how they're going to make rough

00:51:14.540 --> 00:51:15.360
<v Pamela Fox>better for agents.

00:51:16.180 --> 00:51:17.580
<v Pamela Fox>So they may not even like bother.

00:51:17.860 --> 00:51:19.240
<v Pamela Fox>Like, I mean, yes, a couple of humans

00:51:19.640 --> 00:51:21.680
<v Pamela Fox>somewhere are running rough and actually looking at the results.

00:51:22.340 --> 00:51:23.760
<v Pamela Fox>But yeah, they're, you know,

00:51:23.800 --> 00:51:27.340
<v Pamela Fox>so now they're like much more focused on like optimizing the experience for agents.

00:51:27.760 --> 00:51:29.500
<v Michael Kennedy>My agent runs rough a lot.

00:51:29.700 --> 00:51:33.520
<v Michael Kennedy>I don't know about yours, but I told the standing rule for all of

00:51:33.600 --> 00:51:35.000
<v Michael Kennedy>my agents, if they're doing coding,

00:51:35.300 --> 00:51:35.580
<v Michael Kennedy>that is,

00:51:35.860 --> 00:51:37.120
<v Michael Kennedy>is whenever you're done,

00:51:37.420 --> 00:51:38.359
<v Michael Kennedy>you're not done until

00:51:38.380 --> 00:51:39.460
<v Michael Kennedy>Ruff format,

00:51:40.020 --> 00:51:40.300
<v Michael Kennedy>ruff check,

00:51:40.720 --> 00:51:41.600
<v Michael Kennedy>and ty check.

00:51:41.960 --> 00:51:43.800
<v Michael Kennedy>All of those run without warnings,

00:51:43.980 --> 00:51:44.560
<v Michael Kennedy>without errors.

00:51:44.870 --> 00:51:45.300
<v Michael Kennedy>You know what I mean?

00:51:45.780 --> 00:51:47.700
<v Michael Kennedy>Like if you introduce something that breaks those rules,

00:51:48.240 --> 00:51:48.960
<v Michael Kennedy>those are easy to fix.

00:51:49.480 --> 00:51:49.700
<v Pamela Fox>Right.

00:51:49.770 --> 00:51:50.120
<v Pamela Fox>So that,

00:51:50.290 --> 00:51:52.120
<v Pamela Fox>I mean, you can put that in your agents.md.

00:51:52.650 --> 00:51:54.500
<v Pamela Fox>I oftentimes have that in my agents.md.

00:51:54.730 --> 00:51:57.620
<v Pamela Fox>You can even formalize like a lot of these coding agents

00:51:57.840 --> 00:51:58.820
<v Pamela Fox>have what's called like a hook,

00:51:59.360 --> 00:52:00.620
<v Pamela Fox>at least VS Code, we call it a hook.

00:52:01.220 --> 00:52:03.700
<v Pamela Fox>So you can have a hook that runs every single time

00:52:03.880 --> 00:52:05.140
<v Pamela Fox>and make sure that it runs.

00:52:05.140 --> 00:52:07.200
<v Pamela Fox>But a lot of times if you just put it in your agents.md,

00:52:07.400 --> 00:52:09.280
<v Pamela Fox>then that's enough instruction to,

00:52:09.820 --> 00:52:10.200
<v Pamela Fox>you know, to be,

00:52:10.440 --> 00:52:11.700
<v Pamela Fox>to be always taking those actions.

00:52:12.200 --> 00:52:15.780
<v Michael Kennedy>Yeah. If you want to be super old school, I mean, we're talking like 2024 old school,

00:52:16.130 --> 00:52:18.740
<v Michael Kennedy>you could put it in a, a get pre-commit hook really.

00:52:18.870 --> 00:52:19.440
<v Pamela Fox>Oh yeah, of course.

00:52:19.600 --> 00:52:21.100
<v Pamela Fox>That's pre-commit though. Cause we're like,

00:52:21.300 --> 00:52:22.200
<v Pamela Fox>you know, that's even like,

00:52:22.420 --> 00:52:23.440
<v Pamela Fox>like, like later in the process.

00:52:23.640 --> 00:52:25.540
<v Pamela Fox>Right. So it depends how often, right. So you've got hooks,

00:52:26.020 --> 00:52:28.780
<v Pamela Fox>you know, you've got hooks for the agent hooks are,

00:52:28.940 --> 00:52:29.680
<v Pamela Fox>are much more frequent.

00:52:29.920 --> 00:52:29.980
<v Pamela Fox>Right.

00:52:30.240 --> 00:52:30.420
<v Pamela Fox>So,

00:52:30.730 --> 00:52:30.820
<v Pamela Fox>um,

00:52:31.200 --> 00:52:32.420
<v Pamela Fox>you know, we don't know how often you're committing.

00:52:32.790 --> 00:52:32.880
<v Pamela Fox>Um,

00:52:33.180 --> 00:52:34.719
<v Pamela Fox>but you know, if you want it to happen every

00:52:34.740 --> 00:52:38.420
<v Pamela Fox>time I have an agent hook that just every time the agent stops,

00:52:38.860 --> 00:52:39.920
<v Pamela Fox>my kid giggles,

00:52:40.320 --> 00:52:40.940
<v Pamela Fox>it just plays

00:52:40.940 --> 00:52:42.280
<v Pamela Fox>a sound of her giggling.

00:52:42.500 --> 00:52:42.760
<v Pamela Fox>And it's,

00:52:43.020 --> 00:52:44.340
<v Pamela Fox>it's very cute. It just makes,

00:52:44.530 --> 00:52:45.860
<v Pamela Fox>you know, it just fills my life.

00:52:47.560 --> 00:52:47.640
<v Pamela Fox>Yeah.

00:52:47.860 --> 00:52:52.460
<v Pamela Fox>And then it's programmed so that if I leave the machine for more than a minute,

00:52:52.890 --> 00:52:58.900
<v Pamela Fox>then it'll also like kind of giggle harder and tell me the name of the session so that I can

00:52:59.080 --> 00:53:00.240
<v Pamela Fox>decide whether to come back my machine.

00:53:00.400 --> 00:53:02.339
<v Pamela Fox>Right. So that way I can like do my kids laundry while

00:53:02.360 --> 00:53:03.120
<v Pamela Fox>Yeah, yeah,

00:53:03.340 --> 00:53:04.060
<v Michael Kennedy>yeah. That's super cool.

00:53:04.240 --> 00:53:13.320
<v Michael Kennedy>So we're not complete with the CLI conversation. I think one of the things that I would like to encourage people who are building CLIs for agents,

00:53:13.520 --> 00:53:17.860
<v Michael Kennedy>a lot of the things that are interesting about this is it's good for people and it's good for agents.

00:53:17.940 --> 00:53:19.740
<v Michael Kennedy>like because they kind of both use English

00:53:20.240 --> 00:53:21.300
<v Michael Kennedy>or natural spoken language,

00:53:21.860 --> 00:53:24.280
<v Michael Kennedy>it's optimizing for one sort of helps the other a lot of times.

00:53:24.440 --> 00:53:26.660
<v Michael Kennedy>Anyway, so my suggestion about this

00:53:26.690 --> 00:53:30.480
<v Michael Kennedy>is that use a nice CLI framework like Typer or Click

00:53:30.880 --> 00:53:32.140
<v Michael Kennedy>or even if you're using ArcPars,

00:53:32.340 --> 00:53:33.160
<v Michael Kennedy>one of these built-in ones,

00:53:33.480 --> 00:53:36.640
<v Michael Kennedy>make sure there's lots of help documentation and stuff

00:53:36.850 --> 00:53:39.020
<v Michael Kennedy>because then the first thing the agent might do

00:53:39.180 --> 00:53:40.840
<v Michael Kennedy>is your CLI --help.

00:53:41.120 --> 00:53:44.260
<v Michael Kennedy>Okay, I now understand this tool that I have almost completely.

00:53:44.480 --> 00:53:44.800
<v Michael Kennedy>Let's go.

00:53:44.850 --> 00:53:45.380
<v Michael Kennedy>You know what I mean?

00:53:45.680 --> 00:53:46.860
<v Michael Kennedy>You don't have to do much more to teach it.

00:53:46.920 --> 00:53:49.400
<v Pamela Fox>Yeah. And you can even ask the agent like for advice,

00:53:49.700 --> 00:53:49.980
<v Pamela Fox>like, hey,

00:53:50.180 --> 00:53:50.960
<v Pamela Fox>what would help you?

00:53:51.180 --> 00:53:51.260
<v Pamela Fox>Like,

00:53:51.660 --> 00:53:53.080
<v Pamela Fox>what would have helped you achieve that?

00:53:53.360 --> 00:53:53.720
<v Pamela Fox>Right.

00:53:53.960 --> 00:53:54.340
<v Pamela Fox>Like, you know,

00:53:54.520 --> 00:53:55.880
<v Pamela Fox>what were what were the issues with that?

00:53:56.260 --> 00:53:57.220
<v Pamela Fox>Yeah. It's interesting because like,

00:53:57.460 --> 00:53:58.260
<v Pamela Fox>you know, on the one hand,

00:53:58.500 --> 00:54:05.460
<v Pamela Fox>you think it's different to optimize something for an agent versus optimize it for a human. I think that is where some of the current tension is, especially with documentation.

00:54:05.980 --> 00:54:11.740
<v Pamela Fox>trying to figure out is our documentation you know who is the one actually reading it um you know so

00:54:11.800 --> 00:54:16.280
<v Pamela Fox>right now you know we have agents.md and that's definitely targeted agents we have readme.md and

00:54:16.340 --> 00:54:22.020
<v Pamela Fox>a lot of times i end up repeating similar things across both um and yeah and i think just right now

00:54:22.160 --> 00:54:26.800
<v Pamela Fox>it's just currently um attention i think there's some clis that will have kind of like an agent

00:54:27.280 --> 00:54:33.320
<v Pamela Fox>mode that might be like much more verbose for example because agents can actually i think

00:54:33.320 --> 00:54:36.240
<v Pamela Fox>there's actually it goes but sometimes they're like lesser rows because it depends if you're

00:54:36.240 --> 00:54:41.280
<v Pamela Fox>trying to reduce token count or just increase information right because uh you know lms can

00:54:41.460 --> 00:54:46.220
<v Pamela Fox>handle a fair bit of information like they can look through a whole python traceback error right

00:54:46.360 --> 00:54:51.200
<v Pamela Fox>and they can understand it whereas me like you know it takes me a while to get to like the right

00:54:51.500 --> 00:54:56.899
<v Pamela Fox>point in the error um so it's i think i i do find it like really fascinating to try and figure out

00:54:56.920 --> 00:54:57.440
<v Pamela Fox>you know,

00:54:58.040 --> 00:55:01.020
<v Pamela Fox>as the world is moving so much right now,

00:55:01.440 --> 00:55:01.600
<v Pamela Fox>you know,

00:55:01.800 --> 00:55:04.000
<v Pamela Fox>what kind of entity are we

00:55:04.440 --> 00:55:07.000
<v Pamela Fox>optimizing our developer experiences for?

00:55:07.220 --> 00:55:09.700
<v Michael Kennedy>I promised you a GitHub CLI story.

00:55:10.090 --> 00:55:12.600
<v Michael Kennedy>Oh, yeah. So like a year ago or so,

00:55:12.980 --> 00:55:14.480
<v Michael Kennedy>you know, not that long ago,

00:55:14.820 --> 00:55:15.960
<v Michael Kennedy>I was doing something with

00:55:16.160 --> 00:55:20.240
<v Michael Kennedy>Claude and I didn't tell it to go to GitHub. I didn't tell that there was a GitHub.

00:55:20.510 --> 00:55:20.780
<v Michael Kennedy>It just

00:55:20.970 --> 00:55:26.219
<v Michael Kennedy>happened to be in a Git repository and I was working zero intention to go talk to GitHub or

00:55:26.240 --> 00:55:29.080
<v Michael Kennedy>anything and I said, okay, what if we did it this way?

00:55:29.480 --> 00:55:30.280
<v Michael Kennedy>And it says to me,

00:55:30.760 --> 00:55:31.800
<v Michael Kennedy>I'm not logged into VS

00:55:32.000 --> 00:55:32.960
<v Michael Kennedy>code or any of the tools.

00:55:33.500 --> 00:55:36.100
<v Michael Kennedy>I unknowingly was logged into the GitHub CLI.

00:55:36.460 --> 00:55:37.460
<v Michael Kennedy>And I ask it a question. It

00:55:37.460 --> 00:55:37.540
<v Michael Kennedy>goes,

00:55:37.760 --> 00:55:37.940
<v Michael Kennedy>well,

00:55:38.420 --> 00:55:41.960
<v Michael Kennedy>three years ago, you added this GitHub issue and it talked about it this way, but I think

00:55:42.060 --> 00:55:43.100
<v Michael Kennedy>it's changed. Like what,

00:55:43.600 --> 00:55:45.260
<v Michael Kennedy>how, how did you get to my GitHub?

00:55:45.500 --> 00:55:46.720
<v Michael Kennedy>I need to know what is going on here.

00:55:47.120 --> 00:55:52.019
<v Michael Kennedy>And eventually like, oh, it used, I went back and looked like, oh, it used the GitHub CLI to explore

00:55:52.040 --> 00:55:56.480
<v Michael Kennedy>the project and I didn't even tell it to him like what is it doing but in the end I was like you know

00:55:56.580 --> 00:56:00.680
<v Michael Kennedy>what cool it brought something from a long time ago back to the future and like we're we're working

00:56:00.730 --> 00:56:05.260
<v Pamela Fox>on it's good but it freaked me out that that actually reminds me of when I was prepping for

00:56:05.500 --> 00:56:10.900
<v Pamela Fox>today's podcast I asked you know I was like hey you know it's a podcast it's about like Python and MCP

00:56:11.260 --> 00:56:16.320
<v Pamela Fox>and it immediately connected to my emails and my events and it found that you would like attach the

00:56:16.440 --> 00:56:19.899
<v Pamela Fox>Google Doc to the because I didn't know you'd made a Google Doc of what topics we were going to talk

00:56:19.920 --> 00:56:21.180
<v Pamela Fox>about. I had missed that.

00:56:21.430 --> 00:56:23.120
<v Pamela Fox>So it found the Google Doc. I was like, okay,

00:56:23.300 --> 00:56:24.480
<v Pamela Fox>I've made an entire like

00:56:24.600 --> 00:56:27.560
<v Pamela Fox>markdown document prepping you based on the Google Doc. I'm like, what Google Doc?

00:56:28.220 --> 00:56:28.740
<v Pamela Fox>What are you

00:56:28.920 --> 00:56:29.300
<v Pamela Fox>talking about?

00:56:29.720 --> 00:56:30.080
<v Pamela Fox>So,

00:56:30.490 --> 00:56:31.380
<v Pamela Fox>you know, you know these things.

00:56:31.640 --> 00:56:31.820
<v Michael Kennedy>Yeah,

00:56:32.100 --> 00:56:32.180
<v Pamela Fox>right.

00:56:32.360 --> 00:56:32.620
<v Pamela Fox>So like,

00:56:32.960 --> 00:56:33.120
<v Pamela Fox>it's,

00:56:33.420 --> 00:56:33.560
<v Pamela Fox>you know,

00:56:33.930 --> 00:56:35.140
<v Pamela Fox>these agents can be very powerful.

00:56:35.320 --> 00:56:36.080
<v Pamela Fox>They're very good at like,

00:56:36.280 --> 00:56:36.480
<v Pamela Fox>you know,

00:56:36.680 --> 00:56:38.040
<v Pamela Fox>searching through lots of

00:56:38.340 --> 00:56:38.480
<v Pamela Fox>data.

00:56:39.090 --> 00:56:39.220
<v Pamela Fox>I,

00:56:39.440 --> 00:56:43.240
<v Pamela Fox>you know, just even if it's just a read only tool, that's just searching through things

00:56:43.440 --> 00:56:43.580
<v Pamela Fox>like,

00:56:43.780 --> 00:56:47.219
<v Pamela Fox>you know, really, really helpful there because it's not going to get bored after the

00:56:47.240 --> 00:56:52.680
<v Pamela Fox>first page of you know search results right like it can it can keep going um so you know i think it

00:56:52.800 --> 00:56:59.280
<v Pamela Fox>is lowering the barrier to finding information um so you know we don't lose we can't we're not

00:56:59.280 --> 00:57:03.920
<v Pamela Fox>going to lose information as much in the big mess of things because we can have like these

00:57:04.070 --> 00:57:08.860
<v Pamela Fox>really persistent agents that are looking for this so for me from like an information accessibility

00:57:09.200 --> 00:57:13.520
<v Pamela Fox>perspective uh i think that's really good because i've always been very good at like searching the

00:57:13.660 --> 00:57:17.200
<v Pamela Fox>web to the point where even at google people would ask me to search for things because they knew i

00:57:17.220 --> 00:57:22.660
<v Pamela Fox>like really good at googling even at google um and even like in high school i participated in web

00:57:22.730 --> 00:57:27.620
<v Pamela Fox>search competitions because i'm an absolute nerd i wait hold on what search engine do you all use

00:57:27.620 --> 00:57:32.260
<v Michael Kennedy>for that um that wasn't like lycos or anything was it in high school this was for the science

00:57:32.520 --> 00:57:38.560
<v Pamela Fox>olympiad um and that would have been in 2001 so i don't even i think it would have been yahoo

00:57:38.850 --> 00:57:44.339
<v Pamela Fox>at that point okay yeah yeah yeah i was a big yahoo user yeah those were amazing times yeah

00:57:44.660 --> 00:57:45.140
<v Pamela Fox>So, but now,

00:57:45.320 --> 00:57:45.440
<v Pamela Fox>like,

00:57:45.620 --> 00:57:46.360
<v Pamela Fox>you know, now it's fine.

00:57:46.370 --> 00:57:49.000
<v Pamela Fox>I don't have to lord my searching skills anymore.

00:57:49.300 --> 00:57:50.400
<v Pamela Fox>Like now everyone can,

00:57:50.800 --> 00:57:53.320
<v Pamela Fox>you know, use their agent to really efficiently search and find

00:57:53.370 --> 00:57:54.280
<v Pamela Fox>great results, which is good.

00:57:54.460 --> 00:57:58.120
<v Pamela Fox>We're making information universally accessible and useful,

00:57:58.340 --> 00:57:58.620
<v Pamela Fox>which is,

00:57:58.900 --> 00:57:59.360
<v Pamela Fox>I think, the mission

00:57:59.540 --> 00:58:00.320
<v Pamela Fox>statement of Google.

00:58:00.780 --> 00:58:01.840
<v Pamela Fox>But it's, you know, someone's mission.

00:58:02.180 --> 00:58:02.420
<v Michael Kennedy>Yeah,

00:58:02.580 --> 00:58:02.760
<v Michael Kennedy>absolutely.

00:58:03.300 --> 00:58:05.140
<v Michael Kennedy>So final thought on CLIs.

00:58:05.240 --> 00:58:07.160
<v Michael Kennedy>They don't have to necessarily be either or.

00:58:07.390 --> 00:58:10.640
<v Michael Kennedy>I ended up writing a Talk Python podcast and courses CLI,

00:58:11.010 --> 00:58:11.600
<v Michael Kennedy>which is kind of weird.

00:58:12.010 --> 00:58:12.640
<v Michael Kennedy>But if you go,

00:58:12.940 --> 00:58:13.300
<v Michael Kennedy>it's open source.

00:58:13.460 --> 00:58:20.200
<v Michael Kennedy>If you go look at it, all it does is it basically calls the MCP endpoints and the actions are like composition of the tools.

00:58:20.640 --> 00:58:24.120
<v Michael Kennedy>So it really is just a different interface to the MCP server.

00:58:24.180 --> 00:58:29.060
<v Michael Kennedy>But I'm like, the things that want to automate this that don't have access to the MCP server,

00:58:29.280 --> 00:58:31.020
<v Michael Kennedy>like how long would it take to write this?

00:58:31.020 --> 00:58:31.900
<v Michael Kennedy>It wasn't that much work.

00:58:32.120 --> 00:58:32.800
<v Michael Kennedy>So anyway,

00:58:33.020 --> 00:58:33.620
<v Michael Kennedy>you can make them,

00:58:33.960 --> 00:58:36.020
<v Michael Kennedy>they can coexist like cats and dogs living together.

00:58:36.220 --> 00:58:37.120
<v Pamela Fox>Mm-hmm, mm-hmm.

00:58:37.660 --> 00:58:37.820
<v Pamela Fox>Yeah,

00:58:37.980 --> 00:58:38.760
<v Pamela Fox>I mean, that's the thing,

00:58:38.880 --> 00:58:42.940
<v Pamela Fox>is that there's some agents that can connect to MCP servers

00:58:43.360 --> 00:58:44.580
<v Pamela Fox>that can't connect to CLIs,

00:58:44.800 --> 00:58:47.900
<v Pamela Fox>because not all agents have a kind of coding execution environment.

00:58:48.140 --> 00:58:50.120
<v Michael Kennedy>Especially the cloud ones,

00:58:50.500 --> 00:58:50.600
<v Michael Kennedy>right?

00:58:50.620 --> 00:58:53.260
<v Michael Kennedy>If it's just a chat LLN, they're not on your machine.

00:58:53.640 --> 00:58:53.860
<v Pamela Fox>Right,

00:58:54.060 --> 00:58:54.200
<v Pamela Fox>yeah.

00:58:55.060 --> 00:58:58.480
<v Pamela Fox>And then there's other agents that can run CLI commands

00:58:58.620 --> 00:58:59.840
<v Pamela Fox>that for some reason they can't,

00:58:59.880 --> 00:59:01.520
<v Pamela Fox>MCP servers for whatever reason.

00:59:01.700 --> 00:59:04.580
<v Pamela Fox>So it's good to have both.

00:59:04.740 --> 00:59:06.100
<v Michael Kennedy>And it's easier than it used to be.

00:59:06.160 --> 00:59:08.500
<v Michael Kennedy>You should be like, all right, well, who's going to be on the CLI team?

00:59:08.800 --> 00:59:09.340
<v Michael Kennedy>Now it's like,

00:59:09.800 --> 00:59:11.000
<v Michael Kennedy>I know who's going to be on the CLI team.

00:59:12.420 --> 00:59:12.860
<v Pamela Fox>It's all you.

00:59:13.200 --> 00:59:13.580
<v Pamela Fox>It's all me.

00:59:14.160 --> 00:59:15.680
<v Michael Kennedy>Lod, you're writing CLI today.

00:59:15.860 --> 00:59:16.300
<v Michael Kennedy>Let's get going.

00:59:16.640 --> 00:59:17.940
<v Michael Kennedy>Or chat or whatever,

00:59:18.220 --> 00:59:18.400
<v Michael Kennedy>you know,

00:59:18.600 --> 00:59:18.840
<v Michael Kennedy>co-pilot.

00:59:19.180 --> 00:59:20.620
<v Michael Kennedy>Well, this has been super fun.

00:59:20.740 --> 00:59:20.960
<v Michael Kennedy>I mean,

00:59:21.340 --> 00:59:23.300
<v Michael Kennedy>let's close it out with some final thoughts.

00:59:23.800 --> 00:59:24.340
<v Michael Kennedy>What do you think?

00:59:24.700 --> 00:59:24.760
<v Pamela Fox>Yeah,

00:59:24.950 --> 00:59:26.240
<v Pamela Fox>so final thoughts.

00:59:26.410 --> 00:59:29.340
<v Pamela Fox>We haven't really got into the Python aspect of MCP.

00:59:29.540 --> 00:59:35.380
<v Pamela Fox>So things I would tell Python developers is that there is an official Python MCP SDK.

00:59:36.680 --> 00:59:38.460
<v Pamela Fox>and that is to get started.

00:59:38.520 --> 00:59:41.540
<v Pamela Fox>But there's also a fantastic package called Fast MCP,

00:59:41.860 --> 00:59:43.440
<v Pamela Fox>which builds on top of that MCP SDK.

00:59:44.300 --> 00:59:45.220
<v Pamela Fox>And that's from Prefect.

00:59:45.500 --> 00:59:48.880
<v Pamela Fox>And that adds all these like really neat integrations.

00:59:49.180 --> 00:59:51.440
<v Pamela Fox>So I personally always use Fast MCP,

00:59:52.140 --> 00:59:53.240
<v Pamela Fox>but you shouldn't use one of those,

00:59:53.440 --> 00:59:53.520
<v Pamela Fox>right?

00:59:53.580 --> 00:59:55.860
<v Pamela Fox>I was actually looking at a colleague's MCP server today

00:59:56.000 --> 00:59:57.820
<v Pamela Fox>and it was just like handwritten from scratch.

00:59:57.960 --> 00:59:59.420
<v Pamela Fox>And I was like, that's not necessary,

00:59:59.600 --> 00:59:59.700
<v Pamela Fox>right?

00:59:59.780 --> 00:59:59.880
<v Pamela Fox>Like,

01:00:00.200 --> 01:00:02.020
<v Pamela Fox>and if there is an official MCP SDK

01:00:02.520 --> 01:00:04.359
<v Pamela Fox>that passes the specification

01:00:04.380 --> 01:00:07.300
<v Pamela Fox>And then there's Fast MCP, which builds on that.

01:00:07.300 --> 01:00:10.420
<v Pamela Fox>So it also is conformant, but then adds all these additional features.

01:00:11.000 --> 01:00:12.040
<v Pamela Fox>So if you're starting today,

01:00:12.560 --> 01:00:17.480
<v Pamela Fox>you can point your agent at the Fast MCP docs and have it

01:00:18.060 --> 01:00:20.520
<v Pamela Fox>write an MCP server using the latest version of MCP protocol,

01:00:20.800 --> 01:00:22.900
<v Pamela Fox>which is a really nice improved

01:00:23.120 --> 01:00:23.320
<v Pamela Fox>version.

01:00:23.600 --> 01:00:24.280
<v Pamela Fox>It's stateless,

01:00:24.480 --> 01:00:26.240
<v Pamela Fox>so it's more scalable and direction ready.

01:00:27.180 --> 01:00:31.560
<v Pamela Fox>So it is very easy to get started writing your first Python MCP server.

01:00:33.220 --> 01:00:35.480
<v Pamela Fox>And yeah, and there is also a conformance suite.

01:00:35.720 --> 01:00:38.280
<v Pamela Fox>So if for some reason you did write your own MCP server,

01:00:38.780 --> 01:00:41.280
<v Pamela Fox>you should at least run the conformance suite

01:00:41.660 --> 01:00:43.080
<v Pamela Fox>to see that it aligns the spec.

01:00:43.900 --> 01:00:48.320
<v Pamela Fox>But easiest way is start server and just give it a go

01:00:48.320 --> 01:00:49.360
<v Pamela Fox>and maybe just make one for fun

01:00:49.600 --> 01:00:52.900
<v Pamela Fox>just so you can understand how this new protocol works

01:00:53.240 --> 01:00:54.240
<v Pamela Fox>and it'll give you more insight

01:00:54.620 --> 01:00:55.900
<v Pamela Fox>as to when you're using MCP servers

01:00:56.260 --> 01:00:57.960
<v Pamela Fox>with your favorite agent.

01:00:58.240 --> 01:00:59.640
<v Michael Kennedy>Yeah, I don't think I really understood

01:00:59.880 --> 01:01:02.160
<v Michael Kennedy>how MCP servers really worked until I built one.

01:01:02.380 --> 01:01:02.580
<v Pamela Fox>Right.

01:01:02.830 --> 01:01:02.960
<v Pamela Fox>Yeah.

01:01:03.110 --> 01:01:06.080
<v Pamela Fox>Which is like, I mean, that's one of the greatest way of, you know,

01:01:06.450 --> 01:01:07.900
<v Pamela Fox>learning how these things work.

01:01:08.240 --> 01:01:10.180
<v Pamela Fox>I didn't really understand how they worked until I built one,

01:01:10.250 --> 01:01:12.200
<v Pamela Fox>put it in production, and then had an error with it.

01:01:13.619 --> 01:01:15.960
<v Michael Kennedy>Wait, this is not working the way I expected.

01:01:16.140 --> 01:01:16.240
<v Michael Kennedy>Uh-oh.

01:01:16.540 --> 01:01:17.020
<v Michael Kennedy>What's going on?

01:01:17.320 --> 01:01:18.080
<v Pamela Fox>That's when you really,

01:01:18.280 --> 01:01:19.280
<v Pamela Fox>really understand something,

01:01:19.540 --> 01:01:19.600
<v Pamela Fox>right?

01:01:19.800 --> 01:01:21.100
<v Michael Kennedy>That is for sure.

01:01:21.260 --> 01:01:22.680
<v Michael Kennedy>That is for sure. Cool. All right. Well,

01:01:23.040 --> 01:01:23.420
<v Michael Kennedy>Pamela,

01:01:23.640 --> 01:01:24.760
<v Michael Kennedy>thanks for taking the time for being on the show.

01:01:24.960 --> 01:01:26.040
<v Michael Kennedy>Always fun to...

01:01:26.300 --> 01:01:27.080
<v Pamela Fox>All right. Awesome.

01:01:27.340 --> 01:01:28.180
<v Pamela Fox>See you next time.

01:01:28.450 --> 01:01:28.700
<v Michael Kennedy>See ya.

01:01:30.180 --> 01:01:32.460
<v Michael Kennedy>This has been another episode of Talk Python To Me.

01:01:32.840 --> 01:01:33.560
<v Michael Kennedy>Thank you to our sponsors.

01:01:33.790 --> 01:01:35.100
<v Michael Kennedy>Be sure to check out what they're offering.

01:01:35.250 --> 01:01:36.620
<v Michael Kennedy>It really helps support the show.

01:01:37.380 --> 01:01:39.220
<v Michael Kennedy>If you or your team needs to learn Python,

01:01:39.450 --> 01:01:42.840
<v Michael Kennedy>we have over 270 hours of beginner and advanced courses

01:01:43.160 --> 01:01:46.580
<v Michael Kennedy>on topics ranging from complete beginners to async code,

01:01:46.720 --> 01:01:47.420
<v Michael Kennedy>Flask, Django,

01:01:47.820 --> 01:01:49.500
<v Michael Kennedy>HTMX, and even LLMs.

01:01:49.720 --> 01:01:50.200
<v Michael Kennedy>Best of all,

01:01:50.420 --> 01:01:52.000
<v Michael Kennedy>there's no subscription in sight.

01:01:52.540 --> 01:01:54.320
<v Michael Kennedy>Browse the catalog at talkpython.fm.

01:01:55.060 --> 01:01:57.000
<v Michael Kennedy>And if you're not already subscribed to the show

01:01:57.200 --> 01:01:58.420
<v Michael Kennedy>on your favorite podcast player,

01:01:58.620 --> 01:01:59.680
<v Michael Kennedy>what are you waiting for?

01:02:00.140 --> 01:02:02.120
<v Michael Kennedy>Just search for Python in your podcast player.

01:02:02.230 --> 01:02:03.100
<v Michael Kennedy>We should be right at the top.

01:02:03.540 --> 01:02:06.280
<v Michael Kennedy>If you enjoyed that geeky rap song, you can download the full track.

01:02:06.510 --> 01:02:08.340
<v Michael Kennedy>The link is actually in your podcast player show notes.

01:02:09.240 --> 01:02:09.880
<v Michael Kennedy>This is your host,

01:02:10.100 --> 01:02:10.560
<v Michael Kennedy>Michael Kennedy.

01:02:10.960 --> 01:02:12.020
<v Michael Kennedy>Thank you so much for listening.

01:02:12.250 --> 01:02:13.020
<v Michael Kennedy>I really appreciate it.

01:02:13.460 --> 01:02:14.160
<v Michael Kennedy>I'll see you next time.