Today we're launching Jitsu MCP. This is an MCP server that lets your agent run your entire event pipeline directly in Claude Code, Claude Desktop, Cursor, or VS Code. Here's how it works: Add the server: claude mcp add --transport http jitsu Run this prompt: "Route my web stream... into ClickHouse, but only purchase and signup events. Send a test event, check Live Events, and if anything errors, find the function in the path and fix it." Approve the OAuth prompt once, and all 26 tools are live Your agent wires the streams, watches Live Events, and rewrites a function when one of them throws. It runs the fix against a real event before shipping it, so nothing is persisted until it works. Every change lands in the audit log tagged mcp, so you can tell what the agent did from what a person did. It's all MIT licensed, and it runs against your own deployment. Jitsu is one of the fastest growing open-source CDPs with 5,000+ GitHub stars. Capturing events is unlimited and free, and delivery is free up to 200k events a month. Comment Jitsu and I'll DM you the exact prompts I run in Claude Code to build an event pipeline end to end.show more

Here are the Docs, plus the config if you're on Cursor or VS Code instead: Free tier is 200k delivered events a month, so you can point an agent at a real pipeline for absolutely free!

Not on Claude Code? It's in the MCP registry, so you can add it by name. Search Jitsu in your client's MCP directory.

Here's the original Launch Video:

This is the first MCP I’ve seen that actually closes the loop instead of just listing tools. Wiring the stream, checking Live Events, then fixing the function against a real event

Thanks @GesoraMeshack !! How was your experience with the MCP?

Your existing Segment SDKs work unchanged, so trying it doesn't mean rewriting tracking calls. And it's MIT. The MCP server is served at /mcp on your own host, so the agent can be driving a pipeline that never leaves your infrastructure.

This is next level. An MCP that lets your agent wire streams, watch Live Events, rewrite functions and ship only after a real-event test? That's the future of data pipelines. Jitsu just made CDPs actually agentnative.

Thanks so much sanskriti! <3

@sohmehta the audit log tagging is smart. The thing I'd stress-test is drift between what the agent "fixed" autonomously and what your pipeline actually needed changed. Auto-rewrites on live event functions get risky once schema complexity grows.

Hey @ScottShapiroUXD , thanks for this. Nothing gets rewritten onto a live function tho - run_function is a sandbox, candidate code plus an event, you get the result and logs back and nothing is persisted or sent downstream. And you replay the events that actually failed out of the event log, so you're diffing against real traffic. Persisting is a separate destructive call, and there's a read only switch if you want zero writes. Every change lands in the audit log tagged MCP with the diff.

The best part is that Your existing Segment SDKs work unchanged, so trying it doesn't mean rewriting tracking calls. Jitsu is MIT licenced. The MCP server is served at /mcp on your own host, so the agent can be driving a pipeline that never leaves your infrastructure.

Absolutely!

i've seen agents replay production events by mistake, scoped credentials should be default

Simulation before fabrication is standard practice now. Massive game changer

MCP event pipelines could turn analytics agents into operational teammates.

yep!!!

Jitsu

Sent in your DMs!

This is impressive

Thanksss!!!

Jitsu

Sent in your DMs

Hardware testing is the ultimate growth bottleneck. Brilliant move by Antioch

This is a huge step for agent-driven data pipelines 🚀 Being able to build, test, monitor, and fix the entire event pipeline directly from Claude Code is seriously impressive. The MIT license and audit log make this even more compelling. 🔥

thank you 🙌🙌

Really interesting

This is the kind of MCP integration that makes AI agents genuinely useful. Building a workflow is one thing. Letting the agent test it, catch errors, and fix them before anything goes live is a whole different level. Jitsu MCP looks seriously interesting. 🔥

wait you can run the whole pipeline from claude code now?

Yep! You can just point it at your workspace and ask - route this stream into ClickHouse. Claude Code wires it up and then shows you the live events landing. claude mcp add --transport http jitsu We have a generous free tier, would love for you to try it and let us know feedback!

Open source plus real agent control is a strong combo.

💜💜

Agent handling the full pipeline end to end instead of just pieces is the real upgrade ngl

Hey thanks so much! Would love to hear some feedback too :)

jitsu

Check DMs!

mcp access needs scoped credentials before agents can safely touch production pipelines

Yes, and we have a version of this already implemented. Jitsu has a read only switch, you can flip it and every write gets rejected at the boundary before the handler runs, so you can point an agent at production safely. Every tool also carries its own annotation, so a destructive call gets gated by your client.

Jitsu

Audit log tagged mcp is the only reason I’d let an agent touch a live pipeline. The rewrite-on-error loop is where these things usually lie.

if integrating with Claude simplifies workflows, it’ll save devs a ton of time

Stateful test environments are a strong idea. A tool call can be valid in isolation but still wrong for the current workflow state.

my own event pipeline's 'end to end' still means I babysit it end to end. AI agent, tireless, allegedly.

Jitsu + Claude Code + MCP feels like a very natural stack for building data pipelines faster. 🔥

This is the real test for next-generation AI video models

That looks amazing!

the useful part isn't 26 tools. it's that an agent can test and route events without leaving the editor. mcp that maps to a real pipeline (not a toy crud server) is what makes coding agents feel production-shaped.

26 tools available directly to the agent is pretty powerful. This could seriously speed up analytics workflows.

This makes MCP much more useful for developers.

The agent watching Live Events and debugging functions automatically is probably my favorite part of this launch.