ChatML
Cockpit for Claude Code - See everything your agents build
106 followers
Cockpit for Claude Code - See everything your agents build
106 followers
Free, open-source desktop app that gives you full visibility over parallel Claude Code sessions. Watch diffs stream in real-time, review code with AI at three depths, and ship PRs — all from one dashboard. Built entirely by AI.
Hey Product Hunt! 👋
I'm Marcio, co-founder of KnowBe4 (acquired for $4.6B). I built ChatML because I was running 10+ Claude Code agents in parallel and had zero visibility into what they were actually doing.
The CLI is powerful. But when you're orchestrating multiple agents across branches, you're flying blind — tabbing between terminals, losing track of which agent is where, missing code changes, and having no idea what you're spending.
ChatML is the cockpit. One dashboard where you can see every agent, every file change, every diff — in real-time.
Here's the wild part: ChatML was built entirely by AI using ChatML. 900+ pull requests. 60 days. Zero lines of human-written code. 4 programming languages. I was the architect. Claude was the engineer.
What you get:
👁️ Real-time diff streaming — watch code appear as agents write it (sub-50ms)
🔍 AI code review — Quick, Deep & Security depths with inline comments
🚀 PR workflows — review diffs and create PRs without leaving the app
📋 Plan mode — agents propose a plan first, you approve before code gets written
⌨️ Keyboard-first — command palette, customizable shortcuts, no mouse required
💰 Cost visibility — see what each session is spending
🔓 100% open source (GPL-3.0) — no enterprise tier, no catch, no telemetry
Completely free. No accounts. Your code never leaves your machine.
If you use Claude Code, this gives you the visibility and control you're missing. Try it out, and tell us what you'd build!
https://chatml.com | GitHub: https://github.com/chatml/chatml | Twitter: @chatmlx
This is actually pretty cool. The idea of having one place to see what all these AI agents are doing instead of guessing or chasing tabs makes a ton of sense. Feels like things are shifting from doing the work to managing the output, and this kind of sits right in the middle of that. I can see real value here, especially for teams trying to move faster without losing control.
@arthur_alves4 yes. This is exactly why I built it. Handling 10-15 terminal windows was getting too crazy. I needed to move fast and still be able to context switch everything in my head of every agent. We have been using successfully at some of our companies. I rarely open VSCode anymore.
jared.so
How do you handle conflicts when two parallel agents try to modify the same file? Congrats on the launch!