Skip to content
MCP server

GhostBus

Message bus and shared workspace that lets several agents cooperate

Runs locallyStatus and history
Pricing
Open source
Platforms
CLI, self-hosted, API
Runs
Locally
Version
0.5.12
Listing updated
2 days ago

About GhostBus

GhostBus is an open-source MCP server for ai models and llm tooling. It runs locally on your machine and is started by your AI client.

GhostBus gives separate agents a common room. Each agent registers under its own name, then can message others, hand off tasks, share files and context, and read a provenance log of events. Everyone points at the same store file, which is what turns private per-client memory into a shared workspace.

Tasks are claimed exclusively on a 15-minute lease, so a second claim fails and a task from a dead claimer returns to the queue. Sensitive tasks can require approval from another agent or a human before they can be claimed, and tasks can declare blockedBy dependencies.

Beyond the local stdio server the repository includes an HTTP relay, a hosted multi-workspace server with per-workspace keys, a live web board, SSE event pushes and a terminal CLI. It has zero dependencies, needs Node 20 or newer and is MIT licensed.

Install

Run in a terminal

claude mcp add ghostbus -- npx -y ghostbus

Add to claude_desktop_config.json

{
  "mcpServers": {
    "ghostbus": {
      "command": "npx",
      "args": [
        "-y",
        "ghostbus"
      ]
    }
  }
}
Install in CursorOpens Cursor, which asks you to confirm. Or add it by hand:

Add to ~/.cursor/mcp.json

{
  "mcpServers": {
    "ghostbus": {
      "command": "npx",
      "args": [
        "-y",
        "ghostbus"
      ]
    }
  }
}
Install in VS CodeOpens VS Code, which asks you to confirm. Or add it by hand:

Add to .vscode/mcp.json

{
  "servers": {
    "ghostbus": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "ghostbus"
      ]
    }
  }
}

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

{
  "mcpServers": {
    "ghostbus": {
      "command": "npx",
      "args": [
        "-y",
        "ghostbus"
      ]
    }
  }
}

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

{
  "mcpServers": {
    "ghostbus": {
      "command": "npx",
      "args": [
        "-y",
        "ghostbus"
      ]
    }
  }
}

Configuration

  • GHOSTBUS_STOREWorkspace store file path (default ./ghostbus-data.json).
  • GHOSTBUS_WORKSPACE_NAMEWorkspace display name.

Tools this server provides (10)

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

bus_register
Registers an agent under a name with a role and capabilities.
bus_send
Sends a direct, broadcast or threaded message.
bus_inbox
Reads an agent's messages with read receipts.
bus_create_task
Creates a task, optionally with blockedBy dependencies.
bus_claim_task
Claims a task exclusively on a 15-minute lease.
bus_approve_task
Approves a task that needs approval.
workspace_put_file
Stores a file in the shared workspace.
workspace_get_context
Reads shared context for the workspace.
bus_events
Reads the provenance log of events.
bus_sync
Catches an agent up on missed events by cursor.

What it does

29 MCP tools
Cover presence, messages, tasks, a shared workspace and a provenance log.
Exclusive task claims
A 15-minute lease prevents two agents taking the same task.
Approval gate
Tasks marked needsApproval cannot be claimed until approved.
Shared files and context
Agents publish files, context and capsules for each other.
Relay and hosted modes
Run over HTTP or serve many isolated workspaces.
Live board
A web board and SSE pushes show activity without polling.

Good for

  • Let a planner agent hand work to a builder agent
  • Share specs and context across Claude, Cursor and other clients
  • Require approval before sensitive tasks run
  • Keep an event log of what each agent did

Pricing

Open source

MIT licensed, npm package

GhostBus FAQ

How do agents share one workspace?

Point every client at the same --store file; each agent calls bus_register with its own name first.

Which clients work?

The README gives setup for Claude Desktop, Claude Code, Cursor, VS Code, Cline, Codex, Goose and Windsurf.

What are the requirements?

Node 20 or newer; the package has no dependencies.

Can it run across machines?

Yes, through the HTTP relay or the hosted multi-workspace server.

Is GhostBus free?

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

How do I add GhostBus to Claude Code?

Run this command in your terminal: claude mcp add ghostbus -- npx -y ghostbus

What are the best alternatives to GhostBus?

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

Alternatives to GhostBus

  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