Help me test Brain OS

Why this exists

I built a lot with Claude Code and my projects started getting messy. I'd make decisions, change direction, hit blockers, come back a few days later and not remember why I picked one approach over another. I tried Notion. I'm not naturally structured like that.

So I built something for myself. Brain OS is an MCP server that tracks the operational state of a project: decisions and why, what got rejected, what's blocked, what's actually moving, what's quietly drifting. After a while it started showing me things I didn't expect, recurring patterns, "active" projects I hadn't touched in weeks, work I kept calling "next" but never shipped.

That part became more useful than I expected. Now I want to find out whether it's useful for anyone other than me, or whether I built an elaborate tool for my own brain.

Who I'm looking for

• You use Claude Code, Cursor, Zed, or another MCP-compatible AI workflow.
• You have multiple projects or long-running work that gets messy across sessions.
• You can spare ~30 minutes total across the week to try it and reply with honest feedback at the end.
• You're OK with telling me the parts that didn't work, not just the parts that did.

What you'd be testing

Brain OS as it ships today: open source, MIT, local-first. You'd install it (npx brain-os init), connect it to your AI client, use it on real projects for the week, then tell me what happened.

If you want a sense of what the slash commands actually do before you sign up, see the /wrap / /focus / /retro demo on the home page or About.

Verify the code first

Before installing anything from the internet, you should read it. Brain OS is MIT-licensed and small enough to skim in a few minutes — entity store, MCP server, slash commands. No telemetry, no API calls, no hidden behavior.

Sign up