Concord MCP

What it does

Concord MCP provides a shared collaboration layer for AI coding agents. It acts as a local-first workspace where multiple agents (like Claude Code, Codex, and Cursor) can coordinate their efforts, track task progress, and manage explicit ownership of files and modules to avoid collisions.

Tools

• join_workspace: Joins or selects a repository workspace.
• register_agent: Registers agent presence and reads the current roster.
• claim_work: Claims scoped work and records task memory.
• update_task: Updates task state and evidence.
• assign_task / accept_task: Manages explicit task assignment and acceptance.
• offer_handoff / accept_handoff: Coordinates clean handoffs between agents.
• close_task: Records terminal outcomes for a task.

Installation

{
  "mcpServers": {
    "concord": {
      "command": "npx",
      "args": ["-y", "@concord-ai/concord-mcp@latest"]
    }
  }
}

Supported hosts

Confirmed for Claude Code, Codex, and Cursor.