Hey everyone, Tyler here, I built Comity.

Why I built it: AI coding agents got really good, but only alone. On a team, everyone runs their own (one person on Claude Code, another on Codex, someone on Cursor), and none of those agents know the others exist. So two agents build the same thing, one changes a function another relies on and you find out at merge time, and an agent repeats a fix a teammate's agent already tried and gave up on. Then everyone ends up playing messenger, pasting context between people.

Most multi-agent tools run several agents on one person's laptop. Comity is for the other case: different people, different machines, different tools, one team of agents.

What it does:

• Every agent starts with a short briefing on what the rest of the team's agents are building

• Warns before two agents collide on the same code, not at merge time

• Turns a spec, GitHub Issues or a Notion page into one shared task list that every agent claims from, so nobody builds the same thing twice

• Lets agents message each other and hand work off, including what was already tried and didn't work

• Autopilot: step away, and your open agent session keeps taking teammates' work

• Ask my agent: a read-only copy of your agent answers teammates' questions without interrupting it

• The dashboard estimates how much time it saved your team

It plugs into each agent through its own hooks and MCP; setup is three commands. The free plan covers teams of up to 5 developers and 3 repos, with unlimited agents.

Feedback that would help me most:

1. If you're on a team using AI agents: what's the most annoying coordination problem you hit today? Does Comity solve it, or miss it?

2. Setup: did comity init work cleanly with your agent? Anything confusing?
3. Would your team or manager pay for this, and what would they need to see first?

I read every comment and reply to all of them. Brutal honesty welcome.