Tool4lm

🚀 TOOL4LM — Multi-Tool MCP Server for Local LMs

All-in-one tools to supercharge your local or remote LLMs. Search smarter. Calculate faster. Summarize better.

TOOL4LM is an MCP (Model Context Protocol) server written in Node.js/TypeScript that enhances local language models (LM Studio/Ollama/…) with tools: web search, web page fetching & summarization, internal document search & reading, academic research, and calculation. No default API key is required.

✨ Features

• 🔎 Web Search: Search multiple sources (SearXNG + DuckDuckGo HTML), remove duplicates, and return title + URL + snippet.
• 🌐 Web Fetch/Read: Fetch web pages with size/time limits (to prevent SSRF) and extract “readable” text for summarization.
• 📂 Doc Search/Read: Search and read documents in the sandbox directory (txt/md/html/pdf). PDF uses pdf-parse (lazy-load).
• 📚 Scholar Search/Get: Conduct academic research (arXiv + Crossref + Wikipedia) and retrieve metadata by DOI/arXivId.
• 🧮 Calculator: Evaluate expressions with mathjs, supporting precision (rounding).

🧰 Tools & Parameters (Main Name ⇄ Underscore Alias)

openWorldHint is enabled, so the host will accept extra fields if the model accidentally adds them. However, it is recommended to use the parameters as shown in the table for stability.

📦 Installation & Run

npm install
npm run build
npm start

If the server starts successfully, it will print:
[TOOL4LM] started. Sandbox: <sandbox_directory_path>

⚙️ MCP Configuration (LM Studio)

Add the following to the mcp.json file in LM Studio:

{
  "mcpServers": {
    "TOOL4LM": {
      "command": "node",
      "args": ["--enable-source-maps", "C:/path/to/tool4lm/dist/server.js"],
      "type": "mcp",
      "env": {
        "SANDBOX_DIR": "C:/path/to/your_docs",
        "SEARXNG_ENDPOINTS": "https://searx.be/search,https://searx.tiekoetter.com/search",
        "ENGINE_ORDER": "searxng,duckduckgo",
        "LANG_DEFAULT": "vi",
        "REGION_DEFAULT": "vn",
        "MAX_FETCH_BYTES": "1048576",
        "FETCH_TIMEOUT_MS": "8000"
      }
    }
  }
}

On Windows, use / in JSON to avoid escaping \.

💻 Usage Examples

Web search with citation → web.search → web.fetch → web.read → Summarize and attach the URL.
Internal document search → doc.find → doc.read → Summarize and quote passages (include the path).
Academic research → sch.search → sch.get by DOI/arXivId.
Calculation → calc.eval with ^/pow(a,b) (or normalize **).

🌍 Environment Variables

• SANDBOX_DIR — Document directory (default: ./sandbox).
• SEARXNG_ENDPOINTS — Comma-separated list of SearXNG /search endpoints.
• ENGINE_ORDER — For example, searxng,duckduckgo.
• LANG_DEFAULT, REGION_DEFAULT — Suggested language/region for web search.
• MAX_FETCH_BYTES — Maximum page fetch size.
• FETCH_TIMEOUT_MS — Page fetch timeout (ms).

🤝 Contributing

Issues/PRs are highly welcome! If you find a bug or have a suggestion, please open an issue.

📄 License

MIT

☕ Support

If you find this project useful, you can support me via PayPal:

Or click the button below: