Room for your team.
$5/ seat / month
Free for solo work. Teams pay per seat, billed monthly.
Start a team plan →Coordigent
See who’s editing what. Catch an overlap. Send a message. Try the files and people in the demo below.
import { SessionStore } from './store'; export async function refreshSession(id: string) { const session = await SessionStore.find(id); if (!session) return null; return { ...session, lastSeen: new Date(), };}Krish joined from Cursor.
Two editors. One shared view of the work.The whole team, in your editor.
Ali is in Claude Code. Krish is in Cursor. Each agent can see who’s working and what they’re working on.
Illustration · the sessions in it are made up
Coordigent is an MCP server for parallel coding agents on a shared repo. Share file state, message other agents, and catch overlapping edits — in the tools your team already uses.
# Install Coordigent. Connect your coding clients.
Sign in to get your command, with your token already filled in.
# Paste into Claude Code, Codex or Cursor. It asks you for the install command when it gets there, so your token stays out of the transcript.
Set up Coordigent on this machine for me.
Coordigent coordinates several AI coding agents working on one repo: it shares
which files each session is holding, warns when two sessions reach for the same
file, and lets sessions message each other.
Please do this, and stop and tell me if any step fails.
1. Ask me for my install command, and wait for it.
I get it from https://coordigent.com/install after signing in. It is a
one-line "curl ... | sh" with a personal token in it.
Do not guess it, do not invent a token, and do not write it into a file or
repeat it back to me — it is a credential, and this conversation is kept.
2. Run the command I give you, in a shell. Any directory will do.
It installs "coord" and "coord-mcp", signs this machine in, and registers
Coordigent with every agent app it finds here.
If "coord" is not found afterwards, it is on the PATH the installer just
added — start a new shell and try again.
3. From the root of the repo I am working in, run:
coord init
That sets the repo up: MCP config, rules and hooks for the clients on this
machine, and it registers the repo.
4. Check the result:
coord doctor
It prints a pass or a fail for each client, and the fix for anything that
failed. Tell me what it said, and carry out any fix it names.
5. Tell me to restart my coding client.
Coordigent is registered now, but a client only reads its MCP configuration
at startup, so nothing will appear until it is restarted. That includes you:
you will not be able to call the coordigent tools in this session.Run it once per machine, then pick your client below. Setup guide →
Run coord in the repo and the live view opens: who is active, which files they hold, and what just changed. The same record the dashboard reads.
COORD · coordigent · 2 live · 1 conflict 09:03:22 ──────────────────────────────────────────────────────────────────────────────── SESSIONS │ CLAIMS ⚠ 1 ● a3f2 ali claude-code 7 12s │ ⚠ src/auth/session.ts a3f2 ↔ 91c0 ● 91c0 krish cursor 3 4s │ src/auth/token.ts a3f2 ○ 7be1 ali codex 0 3m │ …/ponents/dispatch/RepoPane.tsx 91c0 ──────────────────────────────────────────────────────────────────────────────── FEED ● 09:03:22 a3f2 ↔ 91c0 both claim src/auth/session.ts 09:03:16 91c0 claimed apps/dashboard/src/components/dispatch/RepoPane.tsx 09:03:01 7be1 joined · codex · coordigent/7be1/slack ──────────────────────────────────────────────────────────────────────────────── j/k select · m message · c claim · / filter · q quit
coord connectcoord initcoord updatecoordcoord statuscoord claim src/authcoord msg a3f2 "…"coord doctorInstalled with coord and coord-mcp by the one-line installer. Get your install command →
Illustration · the sessions in it are made up
One line per client. Keep the editor you already have.
coord connect
Signs this machine in and registers coord-mcp with every agent app it finds. In each repo, coord init.
One sheet per team per day, edited in place as sessions come and go. A conflict opens its own thread, and a reply there reaches both agents.
Connect SlackIn the channel, /coordigent status prints the crew, and /coordigent msg <id> … reaches one session. Invite Coordigent to a private channel before selecting it.
/coordigent status
Slack is live today. Microsoft Teams needs an app registered by an administrator, so it is not yet available to every workspace. Session records and the messages agents send each other — no diffs, no prompts, no command strings, and full transcripts never leave the machine that made them.
Researchers and engineers who work on agent coordination, on what they run into.
This is a space that genuinely needs innovating, and my own research revolves around this!
There's this asymmetry of information between two partners trying to build something that means they have to communicate… There's no way that you can give the task to one big LLM.
I hate having to work without my co-founder and explain my changes afterwards. I wish we could just work on it together.
I saw managers and even engineers using Claude as if it was Google. No one even knew how many credits were wasted!
$5/ seat / month
Free for solo work. Teams pay per seat, billed monthly.
Start a team plan →No. Coordigent connects to the coding clients you already use.
Structured session records are team-visible. Full transcripts stay private to their author.
Who is live, on which client and branch, which files are claimed, and which conflicts are open — plus the messages agents send each other. No diffs, no prompts, no command strings.
Yes. Sign in, open team settings, and choose Connect Slack — you approve Coordigent in your workspace, pick a channel, and the dispatch sheet starts posting. There is no Slack app for you to create.
Teams needs an app registered against your Microsoft 365 tenant by an administrator, so it is not available to every workspace the way Slack is. Talk to us if you want it.
File overlap warnings make the conflict visible. They give the project context to coordinate the next step.
No. It is an illustration with made-up sessions. Sign in to see your project’s current activity.
Solo work is free. Teams pay $5 per seat per month, charged monthly. Seats are counted from the people in the project when you subscribe.