Skip to content
MCP server

On Board

Shared project memory and tickets so one coding agent can continue where another stopped

Runs locallyStatus and history
Install
Pricing
Open source
Platforms
macOS, Windows, Linux
Runs
Locally
Version
4.2.0
Listing updated
2 days ago

About On Board

On Board is an open-source MCP server for knowledge and memory. It runs locally on your machine and is started by your AI client.

On Board keeps a .agent-mem folder inside your project that every agent reads and writes, so a different MCP client can continue the work without you re-explaining it. The loop it describes is onboard, read memory, claim work, write progress, hand off. Claude Desktop, Claude Code, Codex, Cursor and Antigravity are named as clients.

Since v4 an agent can park inside memory_wait_for_event and wake as soon as a peer creates a ticket, submits work or delivers a verdict. The site describes a launch trial where a Codex agent and a Claude agent completed a reject, fix and resubmit cycle with no human relay, and the reviewer reproduced a sha256 hash before approving.

Setup is skill-first: npx skills add swisspra/On_Board --skill onboard followed by an /onboard prompt, which reuses one shared runtime and configures one project for one client. It needs Node 22.20 for the skills CLI, Python 3.11 for the setup helper and Python 3.10 with uv or pipx for the runtime. Everything stays local unless you connect other tools. The registry entry requires AGENT_PROJECT_DIR.

Install

Run in a terminal

claude mcp add on_board --env AGENT_PROJECT_DIR='<AGENT_PROJECT_DIR>' -- uvx onboard-memory-mcp

Add to claude_desktop_config.json

{
  "mcpServers": {
    "on_board": {
      "command": "uvx",
      "args": [
        "onboard-memory-mcp"
      ],
      "env": {
        "AGENT_PROJECT_DIR": "<AGENT_PROJECT_DIR>"
      }
    }
  }
}

Add to ~/.cursor/mcp.json

{
  "mcpServers": {
    "on_board": {
      "command": "uvx",
      "args": [
        "onboard-memory-mcp"
      ],
      "env": {
        "AGENT_PROJECT_DIR": "<AGENT_PROJECT_DIR>"
      }
    }
  }
}

Add to .vscode/mcp.json

{
  "servers": {
    "on_board": {
      "type": "stdio",
      "command": "uvx",
      "args": [
        "onboard-memory-mcp"
      ],
      "env": {
        "AGENT_PROJECT_DIR": "<AGENT_PROJECT_DIR>"
      }
    }
  }
}

Add to ~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "on_board": {
      "command": "uvx",
      "args": [
        "onboard-memory-mcp"
      ],
      "env": {
        "AGENT_PROJECT_DIR": "<AGENT_PROJECT_DIR>"
      }
    }
  }
}

Add to your client's MCP settings (most use this shape)

{
  "mcpServers": {
    "on_board": {
      "command": "uvx",
      "args": [
        "onboard-memory-mcp"
      ],
      "env": {
        "AGENT_PROJECT_DIR": "<AGENT_PROJECT_DIR>"
      }
    }
  }
}

Configuration

  • AGENT_PROJECT_DIRAbsolute path to the project that owns the .agent-mem directory.Required

Tools this server provides (1)

Taken from the project documentation; the running server may differ.

memory_wait_for_event
Parks until a peer creates a ticket, submits work or delivers a verdict.

What it does

Project memory folder
Agents share notes in .agent-mem inside the project.
Ticket queue
Agents claim, progress and hand off tasks.
Wake on events
memory_wait_for_event returns when a peer posts work or a verdict.
Handoff history
The next agent sees what the last one did.
Skill-first setup
An /onboard skill writes the project's MCP config.
Local only
Data stays in the project unless you connect more.

Good for

  • Hand work from Claude to Codex without re-explaining
  • Run a reviewer agent that approves or rejects another agent's work
  • Keep a project's decisions across sessions
  • Coordinate several IDE agents on one repository

Pricing

Open source

Prices change. Check swisspra.github.io before you buy.

On Board FAQ

Where is memory stored?

In a local .agent-mem directory in your project, pointed at by AGENT_PROJECT_DIR.

Which clients work?

Claude Desktop, Claude Code, Codex, Cursor, Antigravity and other MCP clients.

What do I need installed?

Node.js 22.20+, Python 3.11+ for the setup helper, and Python 3.10+ with uv or pipx for the runtime.

What does v4 add?

Agents can wake each other through memory_wait_for_event instead of waiting to be told.

Is On Board free?

Yes. On Board is open source, so you can use and self-host it for free.

How do I add On Board to Claude Code?

Run this command in your terminal: claude mcp add on_board --env AGENT_PROJECT_DIR='<AGENT_PROJECT_DIR>' -- uvx onboard-memory-mcp

What are the best alternatives to On Board?

The closest alternatives to On Board are REA, Codebase Memory, GitHub, Serena MCP: the IDE for your agent and XERJ.

Alternatives to On Board

  1. REA

    MCP server

    CLI and MCP server that lets coding agents reverse engineer apps and binaries

    Runs locally
    • Open source
    • 464K GitHub stars
    • Developer tools
  2. Codebase Memory

    MCP server

    Local MCP server that builds a code knowledge graph for AI coding agents

    Runs locally
    • Open source
    • 48.7K GitHub stars
    • Knowledge and memory
  3. GitHub

    MCP server

    Official MCP server connecting AI assistants to GitHub repos, issues and workflows

    Not checked yet
    • Remote
    • Open source
    • 34.5K GitHub stars
    • Developer tools
  4. MCP toolkit giving coding agents symbol-level code retrieval and editing

    Runs locally
    • Open source
    • 31.5K GitHub stars
    • Developer tools
  5. XERJ

    MCP server

    Local search engine for AI agents with cited answers over your code and docs

    Runs locally
    • Open source
    • 19.2K GitHub stars
    • Knowledge and memory
  6. Ouroboros

    MCP server

    Spec-first layer that pins requirements and verifies results of AI coding agent work

    Runs locally
    • Open source
    • 6.4K GitHub stars
    • Developer tools