@maguyva/mcp
Enables code intelligence through semantic, structural, graph, and text search across indexed repositories, allowing AI agents to quickly locate the exact code they need.
README
@maguyva/mcp
Maguyva is a code-intelligence MCP server: semantic, structural, graph, and text search across your indexed repositories, so your agent finds exactly what it needs, fast.
Maguyva is a remote streamable-HTTP MCP server at https://maguyva.tools/mcp. Most modern clients connect to it directly — no local process needed. This package is a stdio bridge for the clients that can't.
Pick your path:
| Your client | Use |
|---|---|
| Claude Code or OpenAI Codex | The maguyva/agent-plugins plugin — bundles the connection, the usage skill, and session-start guidance |
Supports remote MCP with an Authorization header (Cursor, VS Code, Windsurf, Zed, OpenCode, Gemini CLI, …) |
Direct HTTP config — see below |
| stdio-only MCP client | This package |
Whichever path you take, get your API key from the maguyva.ai dashboard, and install the usage skill (see Install the skill) — the skill is what teaches your agent when and how to reach for Maguyva's tools.
Claude Code / Codex
Install the plugin from maguyva/agent-plugins:
/plugin marketplace add maguyva/agent-plugins
/plugin install maguyva@maguyva
export MAGUYVA_API_KEY='<your key>'
The plugin connects over HTTP directly and includes the skill and hooks — nothing else to set up.
Direct HTTP (preferred for everything else)
If your client supports remote MCP servers with a custom Authorization header, connect directly:
{
"mcpServers": {
"maguyva": {
"url": "https://maguyva.tools/mcp",
"headers": {
"Authorization": "Bearer ${MAGUYVA_API_KEY}"
}
}
}
}
Exact config shape varies by client (VS Code uses servers + type: "http", Gemini CLI uses httpUrl, Codex uses TOML, …). The installation docs have a verified copy-pasteable snippet for each client.
This package: the stdio bridge
Only needed when your client speaks stdio but cannot connect to a remote HTTP server directly. It runs locally and forwards between your client's stdio transport and https://maguyva.tools/mcp, reconnecting to the remote without dropping the local pipe.
Requires Node.js. Add to your client's MCP config:
{
"mcpServers": {
"maguyva": {
"command": "npx",
"args": ["-y", "@maguyva/mcp"],
"env": {
"MAGUYVA_API_KEY": "your-api-key"
}
}
}
}
Options: pass --api-key=KEY instead of the env var, and --server=URL (or MAGUYVA_SERVER_URL) to override the endpoint.
Install the skill
Whatever the transport, install the Maguyva skill so your agent knows how to use the tools well — tool selection, when local grep beats indexed search, and troubleshooting:
curl -fsSL https://maguyva.tools/install-skill.sh | sh -s -- --client <client>
--client accepts claude, opencode, codex, cursor, vscode, gemini, cline, or agents (the generic ~/.agents/skills path). The skill is authored to the open Agent Skills standard and lives in maguyva/agent-plugins. Claude Code and Codex plugin users already have it — the plugin bundles it.
License
MIT
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。
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。
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。
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。
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。
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。
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。