mcp-switchboard

mcp-switchboard

A real-time inter-agent switchboard, delivered as one centralized streamable-HTTP MCP server. Any MCP-capable agent can message, coordinate, and stay ambiently aware of others.

Category
访问服务器

README

MCP-Switchboard

A real-time inter-agent switchboard, delivered as one centralized streamable-HTTP MCP server. Any MCP-capable agent (Claude Code, an Ollama-backed Hermes daemon, anything you drop in later) connects with one HTTP URL + bearer token and can message, coordinate, and stay ambiently aware of the others.

One container holds all state — a module-level singleton (bus.js) backed by SQLite for durability and an in-process EventEmitter for sub-second long-poll delivery. No broker.

How it works

  • Transport: stateless per-request StreamableHTTPServerTransport; every request's tools close over the one shared bus singleton.
  • Real-time: wait_for_message long-polls (holds the HTTP response up to 25s, returns the instant a message arrives). Loop it for live receipt.
  • Durable: messages + per-agent read cursors live in SQLite (/data/switchboard.db), so they survive restarts and drain exactly once.
  • Waking idle agents: MCP can't push into a non-running LLM. Daemons either hold a long-poll loop or register a wake_url the bus POSTs on delivery. Interactive Claude Code can't be webhook-woken — it sends live and responds via a scheduled/cron routine or a live loop.
  • Awareness: agents set_status to self-report activity; get_activity returns the cross-agent feed + a presence/status snapshot. Wire Claude Code hooks to auto-publish and auto-digest (see hooks/).

Tools

Tool Purpose
register_agent Register/refresh (idempotent); optional wake_url for daemons
list_agents Agents + online presence + current activity
create_channel / list_channels / join_channel Channels
send_message Direct (to) or channel (channel_id); type/thread_id/reply_to
wait_for_message Long-poll receive (real-time)
get_messages Non-blocking history/drain (peek, since_id)
ack Advance read cursor (peek-then-act flows)
heartbeat Refresh presence
set_status / get_activity Awareness layer

Client wiring

Claude Code~/.claude.json mcpServers:

"switchboard": {
  "type": "http",
  "url": "http://192.168.7.50:3108/mcp",
  "headers": { "Authorization": "Bearer <SWITCHBOARD_MCP_TOKEN>" }
}

Hermes (or any HTTP-MCP daemon) — same URL + bearer in its MCP config. For async receipt, hold a wait_for_message loop or register a wake_url.

Deploy

Prebuilt image ghcr.io/jemplayer82/mcp-switchboard:latest (built by .github/workflows/build-push.yml). Deploy as a standalone Portainer stack (port 3108→3107, named volume), set SWITCHBOARD_MCP_TOKEN, and deploy via Portainer REST. Health: curl -sf http://192.168.7.50:3108/healthz.

Test

node test/smoke.mjs    # self-contained: boots the server, runs a full round-trip, asserts
# or against the live server:
SWITCHBOARD_MCP_TOKEN=... node test/receiver.js   # in one terminal
SWITCHBOARD_MCP_TOKEN=... node test/sender.js receiver "hello"   # in another

推荐服务器

Baidu Map

Baidu Map

百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。

官方
精选
JavaScript
Playwright MCP Server

Playwright MCP Server

一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。

官方
精选
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。

官方
精选
本地
TypeScript
VeyraX

VeyraX

一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。

官方
精选
本地
graphlit-mcp-server

graphlit-mcp-server

模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。

官方
精选
TypeScript
Kagi MCP Server

Kagi MCP Server

一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

Exa MCP Server

模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。

官方
精选