# AgentBrain Handoffs

> AgentBrain Handoffs is an open-source MCP server for developer tools. Exactly-once delivery of handoffs between AI coding agents, with a live status page. It runs locally on your machine and is started by your AI client.

- **Kind:** MCP server
- **Pricing:** Open source
- **Platforms:** CLI, self-hosted
- **Website:** https://huggingface.co/spaces/willykeenan/agentbrain-handoffs
- **Repository:** https://github.com/willykeenan/agentbrain-handoffs
- **MCP Registry name:** io.github.willykeenan/agentbrain-handoffs
- **Categories:** developer tools, productivity
- **Listing updated:** 2026-11-16

## About

The problem it targets is delivery, not messaging: a handoff can get lost, arrive twice or land in a session that is mid-turn. The package sends each handoff as one new turn in the recipient's existing session, waits if that agent is busy, reserves an attempt before writing, and treats a lost reply as uncertain so it is observed rather than resent.

A work contract is a handoff with a title. The recipient returns a result or a blocker, and the sender closes it as accepted, needing revision or blocked, with an optional due time that triggers one non-waking reminder. The README documents a fourteen-value delivery state list, including waiting, busy, held, uncertain, returned and cancelled. Errored turns are resent a bounded number of times, and a busy agent is never interrupted.

It is a local Python package using only the standard library, with a `handoffs` command, a live dashboard and a stdio MCP server. A demo command runs four simulated agents (Planner, Engineer, Reviewer, Writer) on a local page and deletes its data on exit.

## Features

- **Exactly one turn:** Delivers each handoff to the addressed agent once, at the session it had when enrolled.
- **Never interrupts:** A busy recipient keeps its turn and the handoff waits.
- **Bounded retries:** Resends errored turns up to twice and interrupted turns once.
- **Uncertain state:** Observes a possibly delivered write instead of resending it.
- **Work contracts:** Handoffs with titles, returned results, sender decisions and optional reminders.
- **Live dashboard:** Shows agent status, handoffs and a timeline of events.
- **Offline demo:** Simulates four agents without contacting anything real.

## Good for

- Pass a task from a planning agent to an engineering agent without losing it
- Prevent duplicate deliveries when a provider drops a reply
- Track which handoffs are waiting, running or returned
- Close out delegated work as accepted, revised or blocked
- Try the delivery model locally with simulated agents

## Install

### Claude Code (Terminal)

```sh
claude mcp add agentbrain-handoffs -- docker run -i --rm ghcr.io/willykeenan/agentbrain-handoffs:0.1.0
```

### Claude Desktop (claude_desktop_config.json)

```json
{
  "mcpServers": {
    "agentbrain-handoffs": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "ghcr.io/willykeenan/agentbrain-handoffs:0.1.0"
      ]
    }
  }
}
```

### Cursor (~/.cursor/mcp.json)

```json
{
  "mcpServers": {
    "agentbrain-handoffs": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "ghcr.io/willykeenan/agentbrain-handoffs:0.1.0"
      ]
    }
  }
}
```

### VS Code (.vscode/mcp.json)

```json
{
  "servers": {
    "agentbrain-handoffs": {
      "type": "stdio",
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "ghcr.io/willykeenan/agentbrain-handoffs:0.1.0"
      ]
    }
  }
}
```

### Windsurf (~/.codeium/windsurf/mcp_config.json)

```json
{
  "mcpServers": {
    "agentbrain-handoffs": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "ghcr.io/willykeenan/agentbrain-handoffs:0.1.0"
      ]
    }
  }
}
```

### Other clients (your client's MCP settings (most use this shape))

```json
{
  "mcpServers": {
    "agentbrain-handoffs": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "ghcr.io/willykeenan/agentbrain-handoffs:0.1.0"
      ]
    }
  }
}
```

## FAQ

### How do I try it?

Install the package from a clone and run handoffs demo, then open the local page it prints.

### Will a handoff ever interrupt a running agent?

No. A busy recipient keeps its turn and the handoff waits as busy.

### What happens when a reply is lost?

The attempt becomes uncertain and is reconciled by observation, never resent blindly.

### What does it require?

Python 3.9 or newer and only the standard library.

### Is AgentBrain Handoffs free?

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

### How do I add AgentBrain Handoffs to Claude Code?

Run this command in your terminal: claude mcp add agentbrain-handoffs -- docker run -i --rm ghcr.io/willykeenan/agentbrain-handoffs:0.1.0

### What are the best alternatives to AgentBrain Handoffs?

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

## Alternatives

- [REA](https://ai-powered.com/mcp/rea): REA is an open-source MCP server for developer tools.
- [Codebase Memory](https://ai-powered.com/mcp/codebase-memory): Codebase Memory is an open-source MCP server for knowledge and memory.
- [GitHub](https://ai-powered.com/mcp/github-2): GitHub is an open-source MCP server for developer tools.
- [Serena MCP: the IDE for your agent](https://ai-powered.com/mcp/serena-mcp-the-ide-for-your-agent): Serena MCP: the IDE for your agent is an open-source MCP server for developer tools.
- [XERJ](https://ai-powered.com/mcp/xerj): XERJ is an open-source MCP server for knowledge and memory.
- [Ha MCP](https://ai-powered.com/mcp/ha-mcp): Ha MCP is an open-source MCP server for productivity.
- [Ouroboros](https://ai-powered.com/mcp/ouroboros): Ouroboros is an open-source MCP server for developer tools.
- [Claude Code Ultimate Guide](https://ai-powered.com/mcp/claude-code-ultimate-guide): Claude Code Ultimate Guide is an open-source MCP server for ai models and llm tooling.

---
Source: https://ai-powered.com/mcp/agentbrain-handoffs
