treesitter-mcp

🌳 Tree-sitter Code Search MCP Server

A high-precision, multi-language code search and analysis server for AI agents and developers, built on the Model Context Protocol (MCP).

This server leverages the power of Tree-sitter to parse code into concrete syntax trees, enabling structural, syntax-aware queries that are impossible with traditional text-based search. This means zero false positives and a deep, semantic understanding of your codebase.

💡 Why Use This Server?

Capability	Traditional Grep	Tree-sitter Server
Accuracy	Many False Positives	✅ Zero False Positives
Context-Awareness	Low (text-based)	✅ High (syntactic)
Multi-line Patterns	Fragile & Complex	✅ Trivial & Robust
Language Semantics	None	✅ Understands class vs func

✨ Features

• 🧠 Structural Search: Find code patterns based on syntax, not just text.
• 🌐 Multi-Language Support: Ships with support for TypeScript, JavaScript, Python, Go, and Java out of the box.
• 🗺️ High-Level Analysis: Quickly list all functions, classes, or imports in a file to map out its structure.
• ✂️ Contextual Snippets: Surgically extract the full body of a function or class for focused analysis.
• 🏗️ Clean & Modular: Built with a clean architecture for easy extension and maintenance.
• 📊 Token-Aware: Includes middleware to count and log token usage for all tool responses.

🛠️ Available Tools

This server provides a powerful suite of tools for code analysis.

Tool	Description	Primary Use Case
initialize_treesitter_context	Initializes the parser and loads language grammars.	Must be called first. Prepares the server for a session.
list_code_elements_by_kind	Enumerates top-level code constructs (functions, classes, etc.).	Get a high-level "table of contents" for a file.
get_contextual_code_snippets	Retrieves the full containing function or class body for a given code location.	Zoom in on a specific feature's implementation.
structural_code_search	Performs a precise, syntax-aware search for code patterns using a Tree-sitter query.	Find specific, complex code patterns with no false positives.

For detailed technical guidelines, see .clinerules/treesitter-server-guidelines.md.

✨ What's New (as of d957324)

• Integration Testing Guidelines: Added comprehensive guidelines for writing integration tests for new tools. This ensures that all new features are tested robustly and consistently. For more details, see .clinerules/integration-testing-guidelines.md.

🚀 Getting Started

1. Installation

Clone the repository and install the dependencies using pnpm.

git clone <repository-url>
cd tree-sitter-mcp
pnpm install

2. Build

Compile the TypeScript source code.

pnpm run build

The compiled output will be in the build/ directory.

⚙️ Setup for Use

To use this server with an MCP client like Cline, you need to add it to your client's configuration file.

1. Locate your configuration file. For Cline, this is typically found at:

   • macOS: ~/Library/Application Support/Code/User/globalStorage/saoudrizwan.claude-dev/settings/cline_mcp_settings.json

2. Add the server configuration: Open the JSON file and add the following entry to the mcpServers object. Make sure to replace <path-to-this-project> with the absolute path to this project's directory on your machine.

{
  "mcpServers": {
    "... other servers": {},
    "treesitter-code-search": {
      "command": "node",
      "args": [
        "<path-to-this-project>/tree-sitter-mcp/build/index.js"
      ],
      "disabled": false,
      "autoApprove": [
        "initialize_treesitter_context",
        "structural_code_search",
        "list_code_elements_by_kind",
        "get_contextual_code_snippets"
      ]
    }
  }
}

3. Restart your MCP client. The Tree-sitter server should now be available.

🤖 Prompt for AI Agents

To ensure a coding agent effectively uses this server for code analysis tasks, you can use a system prompt or initial instruction like this:

System Prompt:

You have access to a powerful treesitter-code-search MCP server for code analysis. To use it effectively, you must follow this workflow:

1. Initialize: Begin every session by calling initialize_treesitter_context with the languages in the project (e.g., ['typescript', 'python']).
2. Analyze: Call any analysis tool (structural_code_search, list_code_elements_by_kind, etc.) with the absolute path to a file. The server will automatically parse the file on the first request. There is no need to call a separate parsing tool.
3. Explore: Use list_code_elements_by_kind to get a high-level overview of a file's structure (e.g., list all 'function_declaration').
4. Analyze: Use get_contextual_code_snippets to view the full implementation of a specific function or class you've identified.
5. Search: Use structural_code_search with a Tree-sitter query for precise, syntax-aware searches.

Always prefer these structural analysis tools over simple text search for higher accuracy and zero false positives.

---

Happy coding!