🌳 Tree-sitter Analyzer

English | 日本語 | 简体中文

Tree-Sitter-Analyzer is a local-first code context engine for AI-assisted development — combining fast repository retrieval, AST-based structural analysis, and secure MCP integration.

Its job is not just to parse code. Its job is to help humans and AI agents fetch only the code context they actually need, safely, quickly, and with structural precision.

find the right files → find the right matches → extract the right structure → send only the right context

17 languages · Project-boundary security · Claude Desktop / Cursor / Roo Code · CLI + Python API

✨ What's New in v1.10.5

• get_code_outline MCP tool with TOON format: Outline-first navigation delivering 54-56% token reduction vs JSON. Retrieve hierarchical structure first, then fetch only the bodies you need.
• trace_impact MCP tool: Lightweight call site finder using ripgrep — impact analysis without graph database overhead
• Intent-based tool aliases: AI-friendly tool naming (locate_usage, map_structure) makes tool discovery natural for agents
• Analysis session tracking: Audit multi-step SMART workflows with session IDs and operation history
• 23 critical bug fixes: TOON format return structure, default output format, test assertions - project fully operational
• Measured token savings: Real-world testing shows TOON format reduces output size by 54-56% across small/medium/large files
• Enhanced test coverage: 8,470 tests (100% pass), 88.68% coverage (↑8.35% from v1.10.4)
• Cross-platform verified: All tests pass on Ubuntu, Windows, macOS × Python 3.10-3.13

📖 Full Changelog for complete version history.

🎬 See It In Action

Demo GIF coming soon - showcasing AI integration with SMART workflow

🎯 Why Tree-sitter Analyzer

Tree-sitter Analyzer is an open-source, local-first code context engine for helping AI assistants read only what matters in large codebases.

• Minimal context, not whole-file stuffing: retrieve the smallest useful code regions before sending them to AI
• Evidence-based analysis: combine tree-sitter structure with fd and ripgrep to surface relevant files, symbols, and paths
• No heavy preprocessing required: useful on messy repositories where full indexing can be slow, stale, or difficult to maintain

Common Use Cases

• Understand what a very large file or module is doing without loading the entire file into an AI prompt
• Trace business logic, UI handlers, or bug-related code paths across a complex repository
• Narrow AI context for Java and other large codebases before asking for analysis or changes

🚀 5-Minute Quick Start

Prerequisites

# Install uv (required)
# macOS/Linux
curl -LsSf https://astral.sh/uv/install.sh | sh
# Windows PowerShell
powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"

# Install fd + ripgrep (required for search features)
brew install fd ripgrep          # macOS
winget install sharkdp.fd BurntSushi.ripgrep.MSVC  # Windows

📖 Detailed Installation Guide for all platforms.

Verify Installation

uv run tree-sitter-analyzer --show-supported-languages

🤖 AI Integration

Configure your AI assistant to use Tree-sitter Analyzer via MCP protocol.

This works especially well when your assistant struggles with very large files, noisy repository-wide context, or legacy code that is too expensive to load all at once.

Claude Desktop / Cursor / Roo Code

Add to your MCP configuration:

{
  "mcpServers": {
    "tree-sitter-analyzer": {
      "command": "uvx",
      "args": [
        "--from", "tree-sitter-analyzer[mcp]",
        "tree-sitter-analyzer-mcp"
      ],
      "env": {
        "TREE_SITTER_PROJECT_ROOT": "/path/to/your/project",
        "TREE_SITTER_OUTPUT_PATH": "/path/to/output/directory"
      }
    }
  }
}

Configuration file locations:

• Claude Desktop: %APPDATA%\Claude\claude_desktop_config.json (Windows) / ~/Library/Application Support/Claude/claude_desktop_config.json (macOS)
• Cursor: Built-in MCP settings
• Roo Code: MCP configuration

After restart, tell the AI: Please set the project root directory to: /path/to/your/project

📖 MCP Tools Reference for complete API documentation.

💻 Common CLI Commands

Installation

uv add "tree-sitter-analyzer[all,mcp]"  # Full installation

Top 5 Commands

# 1. Analyze file structure
uv run tree-sitter-analyzer examples/BigService.java --table full

# 2. Quick summary
uv run tree-sitter-analyzer examples/BigService.java --summary

# 3. Extract code section
uv run tree-sitter-analyzer examples/BigService.java --partial-read --start-line 93 --end-line 106

# 4. Find files and search content
uv run find-and-grep --roots . --query "class.*Service" --extensions java

# 5. Query specific elements
uv run tree-sitter-analyzer examples/BigService.java --query-key methods --filter "public=true"

╭─────────────────────────────────────────────────────────────╮
│                   BigService.java Analysis                   │
├─────────────────────────────────────────────────────────────┤
│ Total Lines: 1419 | Code: 906 | Comments: 246 | Blank: 267  │
│ Classes: 1 | Methods: 66 | Fields: 9 | Complexity: 5.27 avg │
╰─────────────────────────────────────────────────────────────╯

📖 Complete CLI Reference for all commands and options.

🌍 Supported Languages

📖 Features Documentation for language-specific details.

📊 Features Overview

🏆 Quality & Testing

Metric	Value
Tests	Multi-thousand automated tests
Coverage
Type Safety	100% mypy compliance
Platforms	Windows, macOS, Linux

# Run tests
uv run pytest tests/ -v

# Generate coverage report
uv run pytest tests/ --cov=tree_sitter_analyzer --cov-report=html

🛠️ Development

Setup

git clone https://github.com/aimasteracc/tree-sitter-analyzer.git
cd tree-sitter-analyzer
uv sync --extra all --extra mcp

Quality Checks

uv run pytest tests/ -v                    # Run tests
uv run python check_quality.py --new-code-only  # Quality check
uv run python llm_code_checker.py --check-all   # AI code check

📖 Architecture Guide for system design details.

🤝 Contributing & License

We welcome contributions! See Contributing Guide for development guidelines.

⭐ Support

If this project helps you, please give us a ⭐ on GitHub!

💝 Sponsors

@o93 - Lead Sponsor supporting MCP tool enhancement, test infrastructure, and quality improvements.

💖 Sponsor this project

📄 License

MIT License - see LICENSE file.

🧪 Testing

Test Coverage

Metric	Value
Test Suite	Multi-thousand automated tests across unit, integration, regression, property, benchmark, and compatibility layers
Code Coverage
Type Safety	100% mypy compliance

Running Tests

# Run all tests
uv run pytest tests/ -v

# Run specific test category
uv run pytest tests/unit/ -v              # Unit tests
uv run pytest tests/integration/ -v         # Integration tests
uv run pytest tests/regression/ -m regression  # Regression tests
uv run pytest tests/benchmarks/ -v         # Benchmark tests

# Run with coverage
uv run pytest tests/ --cov=tree_sitter_analyzer --cov-report=html

# Run property-based tests
uv run pytest tests/property/

# Run performance benchmarks
uv run pytest tests/benchmarks/ --benchmark-only

Test Documentation

Test Categories

• Unit Tests: Test individual components in isolation
• Integration Tests: Test component interactions
• Regression Tests: Ensure backward compatibility and format stability
• Property Tests: Use Hypothesis-based invariant checking
• Benchmark Tests: Track performance and regression signals
• Compatibility Tests: Validate cross-version behavior

CI/CD Integration

• Test Coverage Workflow: Automated coverage checks on PRs and pushes
• Regression Tests Workflow: Golden Master validation and format stability checks
• Performance Benchmarks: Daily benchmark runs with trend analysis
• Quality Checks: Automated linting, type checking, and security scanning

Contributing Tests

When contributing new features:

1. Write Tests: Follow the Test Writing Guide
2. Ensure Coverage: Maintain >80% code coverage
3. Run Locally: uv run pytest tests/ -v
4. Check Quality: uv run ruff check . && uv run mypy tree_sitter_analyzer/
5. Update Docs: Document new tests and features

📚 Documentation

🎯 Built for developers working with large codebases and AI assistants

Making every line of code understandable to AI, enabling every project to break through token limitations