gc_mcp
Exposes Grand Central daemon capabilities as MCP tools, wrapping gc_daemon via configurable transports (stdio or HTTP) with strict parameter validation and read-only GitHub tools for issue and repository access.
README
gc_mcp
gc_mcp exposes Grand Central daemon capabilities as MCP tools.
It is a client, not a server of record: every tool call is proxied to a running
gc_daemon. Without one listening on
GC_DAEMON_URL (default http://localhost:4242) the tools will load and then
fail to reach anything.
Install
Point your MCP client at it — no clone or build required:
{
"mcpServers": {
"gc": {
"command": "npx",
"args": ["-y", "@fosferon/gc-mcp"]
}
}
}
Or install it and use the binary directly:
npm install -g @fosferon/gc-mcp
gc-mcp # stdio (default) — the binary is unscoped
GC_MCP_TRANSPORT=streamable-http gc-mcp
From a clone, npm install && npm run build then point the client at
node /path/to/gc_mcp/dist/index.js.
Parameter validation
Every tool rejects unsupported top-level parameters before its callback can
contact gc_daemon. MCP returns its standard isError: true tool result; the
diagnostic names the rejected parameter and lists the parameters registered for
that tool. This makes a misspelled or obsolete option recoverable instead of
silently ignoring it.
This boundary does not alter nested payload contracts: a declared map or object parameter continues to accept the nested values its existing schema allows.
Configuration
| Variable | Default | Purpose |
|---|---|---|
GC_DAEMON_URL |
http://localhost:4242 |
Base URL of the gc_daemon this wraps. |
GC_MCP_TRANSPORT |
stdio |
stdio, streamable-http, or all. |
GC_MCP_HOST |
127.0.0.1 |
Bind host, streamable-http only. |
GC_MCP_PORT |
8765 |
Bind port, streamable-http only. |
GC_MCP_PATH |
/mcp |
HTTP path, streamable-http only. |
GH_TOKEN |
— | GitHub token for the gh_* tools. Falls back to ~/.config/gh-token. |
GH_DEFAULT_REPO |
unset | owner/repo used by the read-only gh_* tools when repo is omitted. |
Why GH_DEFAULT_REPO does not apply to writes
gh_issue_create, gh_issue_edit and gh_issue_comment require an explicit
repo. A write that infers its target from ambient configuration will
eventually file into the wrong tracker, and the caller who omitted the argument
gets no signal that it happened — the operation succeeds, somewhere else.
Reads carry no such consequence, so they may fall back to the environment.
There is deliberately no compiled-in default. A shipped repository name aims every installation's bare calls at whatever tracker the author last worked on.
推荐服务器
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。
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。
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。
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。
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。
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。
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。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
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。