Capture Event Data into Your with Jitsu (YC S20) 🔥 Open-source, compatible with Segment
Joined November 2019
- Tweets172
- Following394
- Followers564
- Likes225
full session with @rohitdotmittal and jitsu's co-founder vlad on YouTube: youtu.be/LiYffwouqJ8?si=fMp_…
Imagine your warehouse goes down for maintenance.
Your product is still generating events. But your warehouse can't ingest them.
What happens to all that data?
This is exactly why Jitsu doesn't send every event directly to the warehouse as an individual INSERT.
As volume grows, thousands of tiny writes can cause problems like Too many parts in ClickHouse or create bottlenecks in Postgres.
In our recent AMA with @sohmehta, Jitsu co-founder Vlad explained how we handle this.
Events first go into Kafka, where they're queued. Jitsu then flushes them to the warehouse in batches.
So even when your warehouse is temporarily unavailable, events stay in Kafka and get loaded once it's back.
The idea is simple: your data pipeline shouldn't depend on your warehouse being available every second.
And whether you run Jitsu on-prem or use Jitsu Cloud, you get the same approach to keeping ingestion moving.
Vlad (our co-founder) told us he barely opens the Jitsu UI anymore.
He just talks to Claude:“What you do is just talking with your agent, let’s say Claude, and basically giving it the directions how to configure things.”
That’s what the MCP server is for.
Claude Code, Cursor, and other agents can create destinations, wire up streams, read Live Events, and edit Functions.
No clicking through screens.
No writing the JavaScript yourself.
Open-source Segment alternative.
Setup: jitsu.com/docs/mcp
full session with @rohitdotmittal, @sohmehta and jitsu's co-founder vlad on YouTube: youtu.be/LiYffwouqJ8
Jitsu is the open-source Segment alternative: collect events from web, app, and server, transform them, and deliver them in real time to your warehouse and tools. MIT-licensed, self-hostable, no MTU pricing.
Our CEO @rohitdotmittal with @sohmehta on why @Heliumventureco is different from private equity:
“We are not spreadsheet buyers.”
Traditional PE projects historical numbers forward and looks for costs to cut.
Helium buys products customers already like and invests to grow them after the deal.
Rohit says Helium has never laid off anyone at a company it acquired.
Since the acquisition we’ve already expanded production self-hosting (Kubernetes-native), added enterprise controls and audit logs, shipped reverse ETL, and added more destinations.
Open-source and self-host remain free. Existing Cloud plans are unchanged.
Customers keep control. The product keeps improving.
If you’re a data/engineering team tired of Segment bills or brittle pipelines, or a founder who built something real that didn’t fit the venture path - this is what a different kind of buyer looks like.
BREAKING NEWS 🚨🚨: Jitsu's co-founder has stopped opening Jitsu UI.
Vladimir Klimontovich co-founded Jitsu as an open-source alternative to Segment. Today he barely touches its dashboard.
"I hardly use any UI. Things I do, I do through MCP."
His setup is just: create the account and billing once.
After that it's just basic prompting. Ask Claude Code or Cursor for a new destination, a stream, a change to a Function, and the agent does it through Jitsu's MCP server and walks you through whatever is left.
What Jitsu agent can actually do there:
- create destinations and wire up streams
- read Live Events to see what is flowing right now
- run a Function against a sample event before it ships
- start connector syncs and pull usage stats
His advice to everyone else: "I suggest you do the same."
Full conversation with @rohitdotmittal and @sohmehta in the comments
Watch the full video unfiltered on YouTube youtu.be/LiYffwouqJ8
raising venture is easy. getting a company acquired is 10x harder - nobody's running after you.
that's @rohitdotmittal for you, YC W16 alum, someone who raised $375Mn in equity + dept, sold his company for millions of $$, and is now the CEO of @heliumventureco and @jitsucom.
he's seen 9 figure deals die at the finish line because the buyer changed CEOs
In the video I speak with Rohit about his journey at YC, raising venture and running his own company to now running his own PE firm and acquiring venture backed companies upto $10M ARR.
We also talk about what makes Jitsu a great product for mid market/enterprise customers and what the future roadmap is going to look like.
Full session's in YouTube. Link's in the first comment.
Super excited for this one
A normal event pipeline only knows the event that just arrived. The history of that user lives somewhere else, so a first-time buyer and a repeat buyer are treated like the exact same thing when data flows to destinations.
Which is exactly why we have Jitsu functions.
A Jitsu function is javascript code that runs on every event before it reaches any destination.
So you and your team can rewrite the event, add fields, or drop it entirely.
One of the helpers you receive is getWarehouse, which runs a SQL query against the warehouse from inside the function.
An Order Completed event can therefore arrive, the function can ask ClickHouse how many orders that user has already placed, write the number onto the event as lifetime_orders, and let the enriched event continue to its destinations - @Meta Conversions API included.
The ad platform receives an event that already knows whether this is the first purchase or seventh.
Functions also receive a key-value store with a 31-day default TTL, fetch for calling external APIs, geo derived from the IP, and a parsed user agent that includes a bot flag.
If you're a CTO/VP of data eng and are fed up of managing complex data pipelines or tired of your $250k Segment bill, DM @sohmehta
jitsu.com/docs/functions
