buzz-mcp-bridge

buzz-mcp-bridge

Bridges Claude Desktop/Code MCP servers into Buzz agents with a per-agent allowlist, exposing allowed tools as a namespaced stdio MCP server.

Category
访问服务器

README

buzz-mcp-bridge

Give Buzz managed agents the MCP servers your Claude Desktop / Claude Code config already has, behind an explicit per-agent allowlist.

The problem

Buzz managed agents run as embedded engines: buzz-acp spawns the runtime (Claude, Goose, Codex) in ACP mode with host-controlled configuration. Your interactive MCP config, the servers you added in Claude Desktop or with claude mcp add -s user, does not inherit. That is deliberate and correct: your user config is your credentialed keyring (billing, admin consoles, your browser), and thirteen semi-trusted agents should not wake up holding it.

The hook Buzz gives you is mcp_command: one command per agent that speaks stdio MCP. This bridge is that command.

What it does

  1. Reads mcpServers from Claude Code user scope (~/.claude.json) and Claude Desktop (claude_desktop_config.json). Code wins name collisions.
  2. Connects only to the servers you explicitly allow for this agent.
  3. Re-exposes their tools as one stdio MCP server, namespaced <server>__<tool>, the same naming Claude Code uses natively.

No allowlist, no startup. Grants are per agent, on purpose.

Install

git clone https://github.com/hachiflow/buzz-mcp-bridge
cd buzz-mcp-bridge && npm install

Write a tiny wrapper per agent (keeps secrets and allowlists out of Buzz's config file):

#!/bin/sh
# my-agent-mcp.sh
exec node /path/to/buzz-mcp-bridge/index.mjs --allow blotato,grafana

Point the agent's mcp_command at the wrapper. Today that field has no UI; edit managed-agents.json in Buzz Desktop's application-support directory:

~/Library/Application Support/xyz.block.buzz.app/agents/managed-agents.json
  -> your agent's entry -> "mcp_command": "/path/to/my-agent-mcp.sh"

The trap that will eat your evening: Buzz Desktop reads its config at app startup and spawns agents from memory. Restarting the agent is not enough. The order is: quit Buzz Desktop fully, edit the file, relaunch. Then verify from outside, never from inside the agent:

ps eww <agent-harness-pid> | tr ' ' '\n' | grep -c YOUR_EXPECTED_VAR

v0.1 limits (deliberate)

  • stdio and streamable-http upstreams with static headers. Interactive OAuth flows are out of scope; use API-key headers for remote servers.
  • Tool lists snapshot at startup. Tools only (no resources/prompts yet).
  • ${VAR} expansion in header values reads the bridge's environment, so secrets can live in files your wrapper sources rather than in any config.

Why we built it

Hachiflow runs its marketing as a Buzz agent (Gary V) in our own hosted hive. He posts to X through Blotato's MCP, granted through exactly this bridge, and nothing else from our keyring. Field notes: https://hachiflow.com/blog

Apache-2.0. Not affiliated with Block; Buzz is their open-source project.

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