What does Penqwin do differently from full file loading?
It sends structural skeletons (imports, exports, signatures, JSDoc) instead of full source code, reducing token context by 80–95%.
Penqwin MCP Server
@sarinmsari
基本信息
关于 Penqwin MCP Server
An AST-based Model Context Protocol (MCP) server that provides token-efficient codebase skeletons to LLM agents (like Cursor, Claude Desktop, and Antigravity).
#context-router#reduce-llm-token#cut-llm-cost
Penqwin MCP Server 配置
使用下面的配置,将此服务器添加到你的 MCP 客户端。
{
"mcpServers": {
"penqwin": {
"command": "npx",
"args": [
"-y",
"@penqwin/mcp"
],
"env": {
"PENQWIN_API_KEY": "YOUR PENQWIN API KEY",
"PENQWIN_ORG_ID": "YOUR PENQWIN ORG ID",
"PENQWIN_REPO": "YOUR GITHUB REPO URL",
"PENQWIN_API_URL": "https://app.penqwin.com"
}
}
}
}Penqwin MCP Server 工具
5Returns a compact table-of-contents for the repository 'YOUR GITHUB REPO URL'. Lists all tracked source files with their exported symbol names. ALWAYS call this FIRST before any other tool to understand the repository structure. Use the file paths returned here as input to get_folder_skeleton or get_file_skeleton. Cost: ~10-20 tokens per file — very cheap.
Returns compact AST skeletons for all source files under a given folder path prefix. Use this to understand a module or feature area without reading raw source files. Skeletons include: exports, function signatures, type definitions, and doc comments. Cost: ~50 tokens per file — much cheaper than raw source code. Tip: call get_repo_index first to discover valid folder paths.
Returns the AST skeleton for a single specific source file. The skeleton includes: all exports with signatures, imports, class members, and doc comments. Use this when you need the details of one specific file after narrowing down from get_repo_index. For multiple related files, prefer get_folder_skeleton — it is one round trip.
Searches the entire repository for files that export a specific symbol name. Use this to find where a function, class, type, or interface is defined. Returns: file path, language, kind (function/class/type/etc.), signature, and doc comment. Example: search for 'createClient' to find all files that export a function by that name.
Returns aggregate statistics for the repository 'YOUR GITHUB REPO URL'. Includes: total file count and a breakdown by programming language. Use this to understand the tech stack and scale of the codebase at a glance.
概览
What is Penqwin?
An AST-based Model Context Protocol (MCP) server that provides token-efficient codebase skeletons (imports, exports, signatures, JSDoc comments) to LLM agents like Cursor, Claude Desktop, and Antigravity. It reduces token context sizes by 80% to 95% during codebase exploration and navigation.
How to use Penqwin?
Install via npm install @penqwin/mcp and build with npm run build. Configure environment variables (PENQWIN_API_KEY, PENQWIN_ORG_ID, PENQWIN_REPO, PENQWIN_API_URL) and integrate with your IDE using npx or by pointing to the local build. Run node dist/index.js to start the stdio server.
Key features of Penqwin
- Five MCP tools:
get_repo_index,get_folder_skeleton,get_file_skeleton,search_symbols,get_repo_stats. - AST-based extraction reduces token context by 80–95%.
- Requires Node.js v18+ and an active
eng-docbackend server. - Configurable entirely via environment variables.
- Integrates with Cursor, Claude Desktop, and Antigravity/Gemini Code Assistant.
Use cases of Penqwin
- Quickly understand a repository’s structure without loading full source files.
- Search for exported symbols (classes, functions, types) across the codebase.
- Retrieve structural skeletons of a single file with signatures, parameters, and JSDoc.
- Navigate codebase with a compact table of contents (~10–20 tokens per file).
常见问题
What are the runtime requirements?
Node.js v18+ and an active eng-doc backend server with a valid API key generated from the eng-doc platform.
How do I configure Penqwin?
Set environment variables `PENQWIN_API_KEY`, `PENQWIN_ORG_ID`, `PENQWIN_REPO`, and `PENQWIN_API_URL`.
What transport does Penqwin use?
It uses stdio transport for MCP integration.
Which IDEs are supported?
Cursor, Claude Desktop, and Antigravity/Gemini Code Assistant.
AI 与智能体 分类下的更多 MCP 服务器
superbot
EZO Labssuperbot is your own AI agent: anything you can do on your computer, it can do, fast. Its MCP server gives other agents one tool for errands: fax, calls, texts, Gmail drafts, calendar, web search. Free in your browser, n
Lumify Sports Intelligence
LumifyAgent-ready sports intelligence API: live scores, odds, line movement, public betting splits, and explainable bet confidence — via 16 MCP tools at https://lumify.ai/mcp. Get a free key instantly — no signup, email, or ca
Melaya
Melaya Labs**Melaya gives an AI assistant hands.** It is a remote MCP server, so there is nothing to install and no API key to configure — one URL works in Claude, ChatGPT, Le Chat, Cursor and any MCP client. ### Your Android phon
Hostinger
hostingerModel Context Protocol (MCP) server for Hostinger API. If you don't want to install or run anything locally, connect directly to Hostinger's hosted MCP server: This opens a browser window to authorize via OAuth. Once app
148
Pocket Network Agentic Marketplace: Pay-per-Call APIs
Pocket Network FoundationMCP server for the Pocket Network agentic marketplace: search our catalog of data services and utilities and call them, paying per request in USDC on Base with x402. No account or API key. Search and describe are free; p
评论