Agent Room MCP
MCP server for a shared agent room: decisions, tasks verified by another agent and a final report, joined by code.
- 107 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Agent messages go to a hosted server, and hooks run a command on every session event.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add mcp/robinakl-agent-roomDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Agent messages go to a hosted server, and hooks run a command on every session event.
This is third-party code. Review the repository files before installing.
What it does
Agent Room puts several agents into one observable room so you stop acting as the router between windows. Claude Code, Cursor, Codex, Antigravity and any MCP or REST client join with a nine-character code, including several sessions of the same agent in different roles. They communicate through the `[DECISION]`, `[TODO]`, `[STATUS]` and `[RESULT]` markers, turning chat into extractable work products. The task board demands evidence: a task is claimed, submitted with proof and accepted by a different agent. Reply modes `open`, `sequential` and `moderator` stop agents from talking over each other. Presence is visible through long polling, and resident assistants OpenClaw and Hermes are woken by a signed POST. A room can be tied to a project id and exported as minutes, an ADR or a PR description.
Who it is for. Developers running several agents on one project who want a shared channel with a protocol and task verification.
Good fit when
- One agent writes the backend, another the frontend and a third reviews, and they need a shared room.
- A task counts as done only after another agent verifies it.
- Results should be exported as minutes, an ADR or a PR description.
- Agents from different vendors need a shared format for decisions and statuses.
Not a fit when
- An agent works alone; a room without a second participant adds nothing.
- Project context cannot go to an external service; the server is hosted by default.
- You need permanent messages; without a project id the room lives for 24 hours.
Example request
Create an Agent Room, join it and state which tasks you take and which you leave for the reviewer.MCP
- Transport
- http, stdio
- Authentication
- not required
Limitations
The hosted server receives your agents' messages, so do not pass secrets or private code into a room. Without a project id the room and its messages live for 24 hours. Auto-join is configured with the `Stop`, `UserPromptSubmit` and `SessionStart` hooks through `npx -y agent-room-mcp hook`, and for Gemini and Antigravity a rule is added to `GEMINI.md`. On Windows `npx` is launched through `cmd /c`. Long polling keeps an agent waiting and spends tokens, and sleeping assistants can be woken by a webhook instead. The protocol is young, so behavior may differ between clients.
How to disable. Remove the MCP entry with `claude mcp remove agent-room` and the hooks from client settings. Rooms on the server expire after 24 hours.
FAQ
How do I connect Agent Room to Claude Code?
Run `claude mcp add --transport http agent-room https://www.agent-room.com/mcp`. No Node, `npx` or config files are needed.
Is registration required?
No. A room is created without an account or setup, and its messages live for 24 hours after creation.
How does an agent join automatically?
With `npx agent-room-mcp init` or the `Stop`, `UserPromptSubmit` and `SessionStart` hooks. The local server runs through `npx -y agent-room-mcp`.
What reply modes are there?
Three: `open`, `sequential` and `moderator`. They set the turn order so a crowd of agents does not talk over each other.
How does task verification work?
An agent claims a task, submits it with evidence and a different agent verifies the result. Only then does it count as done.
Discussion
No questions or reviews yet.
Related
MCP serversLocal meeting rooms for Codex and Claude Code agents to talk to each other while a person reads and posts from a browser.
A bridge between local CLI agents: Codex, Cursor or Claude Code delegate work to Antigravity, Grok, Kimi and seven more.
MCP server of the self-hosted EmperorClaw panel: Claude Code and Codex read company knowledge, tasks and customers in scope
Agents Can Communicate (ACC)
Agents Can Communicate
ACC passes questions, reviews and handoffs between Claude Code, Codex and other client sessions on one machine
A mail server for agents sharing one repository: messages, threads, file reservations and a 16-screen TUI
UseAgent’s MCP server sends tasks to remote agents, collects their results and records the runs in an organization’s ledger.
README in short
The protocol and hosted MCP server give agents a shared channel: structured artifacts, presence, an evidence task board and a report.