Lets Codex, Claude Code, Cursor, Hermes Agent and OpenClaw share one encrypted local memory, hand off tasks, and review each other's work.
Get-Concord-AI/concord-mcp
An open-source MCP server that lets coding agents like Claude Code, Codex and Cursor message each other, claim files and share task state locally.
- Stars
- 989
- Forks
- 57
- Open issues
- 38
- Last commit
- yesterday
What is it?
Concord MCP is a local-first coordination layer for AI coding agents running in different harnesses. Agents can register presence, send direct prompts and replies to each other, claim files or modules before editing, and detect overlapping claims that would otherwise cause conflicting changes.
The shared state lives in a SQLite database inside the repository's .concord/ directory, alongside generated handoff and review-packet files. Ownership changes such as assignment, handoff offers and acceptance are versioned and recorded in an append-only history, so an agent can pick up a task with its context and evidence instead of reconstructing it from chat logs or diffs.
It is exposed as a single MCP server with a small set of tools, and it also ships a CLI with commands for status, task listing, overlap checks for hooks or CI, and a live terminal dashboard. Setup writes the MCP registration and instructions for supported clients.
Key features
- 01Direct prompts and replies between agents across supported coding clients
- 02Claims files and modules before work begins and flags overlaps
- 03Versioned task ownership with explicit assignment and handoff acceptance
- 04Local SQLite store with generated handoff and review-packet files
- 05CLI commands and a terminal dashboard for inspecting agents and tasks
Use cases
- Running Claude Code and Codex on the same repository without them editing the same files
- Having one agent hand a task and its review evidence to another agent
- Blocking conflicting edits with a pre-commit hook or CI check
- Monitoring which agents are active and which claims have gone stale
Quick start
- 1
Install the package globally with npm
npm install -g @concord-ai/concord-mcp - 2
Go to your repository and run setup
cd /path/to/your/repository - 3
Run concord setup in the repository
concord setup - 4
Restart your agent clients
Things to know
- Live delivery depends on the receiving harness and session state, and some integrations only leave durable pull messages.
- Setup writes config into several agent client files, so it touches more of your environment than a typical CLI tool.
In collections
Topics
FAQ
- What is concord-mcp?
- An open-source MCP server that lets coding agents like Claude Code, Codex and Cursor message each other, claim files and share task state locally. Concord MCP is a local-first coordination layer for AI coding agents running in different harnesses. Agents can register presence, send direct prompts and replies to each other, claim files or modules before editing, and detect overlapping claims that would otherwise cause conflicting changes.
- Is concord-mcp free to use? What license does it have?
- concord-mcp is open source under the MIT license. Check the license terms before using it in commercial products.
- Is concord-mcp still maintained?
- Actively maintained. The last commit to concord-mcp was on Nov 6, 2026.
- What can I use concord-mcp for?
- Running Claude Code and Codex on the same repository without them editing the same files Having one agent hand a task and its review evidence to another agent Blocking conflicting edits with a pre-commit hook or CI check Monitoring which agents are active and which claims have gone stale
- How do I install concord-mcp?
- Install the package globally with npm: npm install -g @concord-ai/concord-mcp Go to your repository and run setup: cd /path/to/your/repository Run concord setup in the repository: concord setup Restart your agent clients
- Where can I get concord-mcp?
- The source code and installation instructions are on GitHub: https://github.com/Get-Concord-AI/concord-mcp
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
Configures AI coding agents like Claude Code, Cursor, OpenCode and Codex with persistent memory, shared workflows, skills and MCP servers.
A CLI and workflow engine that runs coding agents like Claude Code, Codex and Copilot through configurable review, repair and test gates before work is accepted.
A plugin that lets Claude Code and OpenAI Codex delegate tasks to Google's Antigravity CLI (agy) running Gemini models.
Local-first desktop workbench that runs multiple coding agents in parallel, each in its own git worktree, with diffs, editor, terminals and PR actions in one window.
Claude Code plugins that run planned, approved workflows for implementing, designing, reviewing and verifying code changes.