We're having conversations with the team about building Prefect. Whatever comes up. This time with Engineers Nate Nowack (@zzstoatzz) and Alex Streed (@desertaxle). Designing for coding agents, the Prefect MCP server and the Pydantic AI integration. www.youtube.com/watch?v=qTT1...

oh hey its my bio! » uvx pdsx -r zzstoatzz.io \ ls app.bsky.actor.profile \ -o json \ | jq -r '.[0].description' yooper → umich → chicago open-source dev tools @prefect.io ‣ zzstoatzz.io ‣ blog.zzstoatzz.io ◦ status.zzstoatzz.io ‣ github.com/zzstoatzz ‣ tangled.sh/@zzstoatzz.io

Kudos again to the @prefect.io crew on doing documentation right for the AI era. I love everything about this -- mcp server, llms.txt and copy as markdown shortcut behavior

i made a @nooki.me MCP community and posted an introduction to MCP nooki.me/post/an-intr...

This is the most compelling argument I've seen thus far for MCP.

In a recent post, @chris.blue argues that MCP is an unnecessary and temporary kludge on the decline and facing an uncertain future. IMO it standardizes how agents have stateful, identity-aware conversations. aaazzam.substack.com/p/the-apolog...

My new content strategy: 1. Be trending 2. Don’t be not trending

As an open-source maintainer, one of the hardest things is saying "no" to well-designed features that are simply not a good fit. With LLMs able to generate volumes of code without context, the maintainer's role as a thoughtful curator has never been more important. www.jlowin.dev/blog/oss-mai...

Ok we get it... you really, really, really like easy auth integrations.

this time is different

"This issue appears to be a duplicate... of itself." Tell me more about this "AGI" github.com/jlowin/fastm...

Imagine if we'd invested in creating this level of clear, concise onboarding instructions for all our human developers all this time: github.com/jlowin/fastm... Maybe the bots will make us better humans kudos to jlowin.bsky.social and @prefect.io crew

$ head -n 1 CLAUDE.​md > If you ever say "You're absolutely right", you will be fined $200.

fastmcp.cloud is here! 🚀 FastMCP made it easy to build MCP servers. FastMCP Cloud makes it just as easy to deploy them. It's the MCP-native platform you've been waiting for, and it just works!

This is being posted by an MCP client connected to an AT Protocol server running on FastMCP Cloud - enabling programmatic interaction with Bluesky through the Model Context Protocol!

🔥 FastMCP Cloud is here We built FastMCP to simplify MCP servers. Now deployment is instant too. Push repo → Live endpoint in <1min 📕 Learn about FastMCP Cloud-> prefect.io/blog/acceler... 🚀 Deploy for free -> fastmcp.cloud

"You're absolutely right!"

Ladies and gentlemen, I give you: Opus 4.1

Interesting things shipping in FastMCP 2.12...

You call that a commit message? THIS is a commit message.

How does a CLI command to transfer resources between backends relate to Kahn's algorithm and topological sorting? How can the looming threat of an LLM character takedown keep us accountable? These are questions that can only be answered in one place: dev-log.prefect.io

In the near future, “self-driving” is going to mean that you drive yourself

A week maintaining an active open-source project can make you laugh, cry, or reenable spacebar heating to avoid breaking changes. We're starting a dev log to share the good, the bad and the ugly decision making that goes into every week's release of Prefect: dev-log.prefect.io/so-a-dev-log...

reviewing PRs in 2025

I can't emphasize this👇 too much. Our job is not to build the most sparkly "innovative" pie-in-the-sky "unicorn" idea. Maybe one in 10,000 of those take off, and I'm being conservative. Instead, address the pain. 1/3

The adult version of watching the oxygen bar run out in Sonic

One day Claude will actually read CLAUDE.​md, but today is not that day.

The tragedy of venture capital: you raise it by pitching the smartest, most innovative ideas you can imagine. But you win customers by solving their dumbest, most annoying problems. That misalignment kills so many startups. So glad Prefect is now funded by the people who actually use it.

It's hard to overstate what Prefect 3 did for adoption. Usage has now increased nearly 10x since GA less than a year ago... and the rate is accelerating.

I'll believe in AGI when my iPhone can filter out political spam text messages. But not before.

2018 AI demo: book a reservation 2019 AI demo: book a reservation 2020 AI demo: book a reservation 2021 AI demo: book a reservation 2022 AI demo: book a reservation 2023 AI demo: book a reservation 2024 AI demo: book a reservation 2025 AI demo: book a reservation with agents

I was honored to open PayPal's first-ever AI Summit! Watching a company at this scale commit to radical transformation is astounding and inspiring, and I got to see this not only from the keynote stage but in hours of conversations with the AI, LLM, and MCP teams after.

Thrilled to share four milestones we hit in July: 🚀 Record‑breaking quarter: we just posted the strongest in company history. 🏁 Profitable: we’ll affirm this milestone in the next few weeks.

I almost didn’t make cyclopts a fastmcp dependency because it didn’t have a lot of downloads. Problem solved, I guess?

2024: No wifi on the flight? Guess I'll do some coding 2025: No wifi on the flight? Guess I can't do any coding.

🛑 Stop converting your REST APIs to MCP. I know that slick OpenAPI-to-MCP converter feels like magic, but it's poisoning your agent. I'm telling you this as the person who built the most popular one. Here’s why we need to rethink agent-first API design: www.jlowin.dev/blog/stop-co...

“Two months ago we launched self-serve pricing at @prefect.io. Today is Tuesday and it's already our 3rd largest week of sign-ups ever 📈” - @cicdw.dev www.linkedin.com/posts/whitec...

It's all fun and games until an AI vibe-codes a 42,000 line PR and another AI vibe-reviews it as "impressive and substantial." This is the way the world ends, not with a bang but a "LGTM!" github.com/google-gemin...

Sure, I could use AI video to sell something... or I could just sit here and watch the virtual starling murmurations fly by

Git worktrees are the recommended way to run Claude Code in parallel, but managing them is kind of tedious... so I built claude-wt! A simple utility to spin up and manage Claude in isolated worktrees. No install needed: uvx claude-wt new

"Cyclopts is what you thought Typer was." Just refactored the FastMCP CLI and I have to say I agree with this line from the Cyclopts docs.

We're big fans of @mintlify but there's no great way to create Python SDK docs automatically. Until now! @alternatebuild.dev created mdxify to automatically generate beautiful docs from code, and it works great. Check out gofastmcp.com/python-sdk and docs.prefect.io/v3/api-ref/p... to see!

A common question I hear when teams start building for AI is, “Why can’t agents just use our REST API instead of MCP?” My response is usually the same: “Why can’t you just use your REST API instead of a UI?” www.jlowin.dev/blog/curatio...

@alternatebuild.dev wrote up his journey to build an atproto (Bluesky) MCP server, including all the code you need to let your AI post! Lots of interesting details on optimizing the API design. blog.alternatebuild.dev/an-atproto-m...

oh nice, i’ve been wanting this

Introducing MCP-native middleware, as easy as 1-2-FastMCP. With FastMCP 2.9, you can now add cross-cutting logic like auth and logging with high-level hooks that understand your tools and resources. See how it works: www.jlowin.dev/blog/fastmcp...

Elicitation is a new MCP feature that I'm really excited about. When Claude Code asks you "do you want to create this file? yes / no / no and tell Claude why", that's elicitation. It lets programmatic actions reach back into natural-language-land to get 👍/👎 approvals or details. In FastMCP soon!

If you ask Claude Code to read any of the OpenAI docs it hits a 403. 🤷‍♂️

This is succinct and well written. As backend engineers we are now designing APIs for humans, software, and agents. Different concerns for different clients.