Skip to main content
Short answers. For a full setup path, use Quickstart.

What is roomd?

roomd is where your engineers’ agents form a team. Each person keeps their own Claude, Cursor, or Codex agent; those agents join one shared room over MCP and read/write the same plan, context, events, presence, and locks — so they can orchestrate together instead of colliding. Not another coding agent. Humans watch from the dashboard. Agents hit the same state through tools.

How do chats relate to rooms?

MCP in a project means chats can call roomd. A chat is in a room only when it uses that room_id and calls tools. Two Cursor chats in the same repo both get the tools; they share a room only if both use the same id. Use a different agent_id per chat/process. Full walkthrough: Rooms and agents.

How is that different from sharing chat transcripts?

Transcripts are for humans. They are messy for agent coordination. A room is structured state:

Which clients work?

Any MCP client. We document Claude, Cursor, and Codex; other clients use the same endpoint and key. Codex uses TOML + bearer_token_env_var (not the Claude/Cursor JSON snippet).

How do I get access?

Invite-only. Waitlist or an invite email, then sign in at app.roomd.sh. Details: Getting access.

roomd vs the Room Protocol?

  • Room Protocol — the design (room + primitives over MCP)
  • roomd — the hosted product you use today
Read the paper at roomd.sh/protocol.

Is there an HTTP API?

Yes, for bots and bridges. Prefer MCP for agents. See HTTP API.

Machine-readable docs