ChunkHound
chunkhound/chunkhound
Busca e análise semântica de código com IA, 100% local
Ver no GitHubMCP server que fornece inteligência semântica sobre codebases locais, usando o algoritmo cAST para chunking inteligente de código. Permite buscas semânticas em linguagem natural e por regex, suportando 32 linguagens de programação e configuração, mantendo tudo processado localmente na sua máquina.
Instalação
npx convez-ai chunkhound/chunkhound1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Pesquisar padrões arquiteturais e conhecimento institucional em codebases grandes
- Descobrir relacionamentos entre componentes de código via buscas multi-hop semânticas
- Integrar análise de código em fluxos com Claude via MCP
Veredito

ChunkHound é uma ferramenta sólida e inovadora para análise semântica de código que se integra nativamente com Claude via MCP. A abordagem local-first e o algoritmo cAST com fundamentação em pesquisa oferecem vantagem real sobre buscas convencionais em IDEs.
README
A tradução para o português deste README ainda está na fila.
Your AI assistant searches code but doesn't understand it. ChunkHound researches your codebase—extracting architecture, patterns, and institutional knowledge at any scale. Integrates via MCP.
Features
- cAST Algorithm - Research-backed semantic code chunking
- Multi-Hop Semantic Search - Discovers interconnected code relationships beyond direct matches
- Semantic search - Natural language queries like "find authentication code"
- Regex search - Pattern matching without API keys
- Local-first - Your code stays on your machine
- 32 languages with structured parsing
- Programming (via Tree-sitter): Python, JavaScript, TypeScript, JSX, TSX, Java, Kotlin, Groovy, C, C++, C#, Go, Rust, Haskell, Swift, Bash, MATLAB, Makefile, Objective-C, PHP, Dart, Lua, Vue, Svelte, Zig
- Configuration: JSON, YAML, TOML, HCL, Markdown
- Text-based (custom parsers): Text files, PDF
- MCP integration - Works with Claude, VS Code, Cursor, Windsurf, Zed, etc
- Real-time indexing - Automatic file watching, smart diffs, seamless branch switching, and explicit backend selection (
watchdog,watchman,polling)
Documentation
Visit chunkhound.ai for documentation:
Requirements
- Python 3.10+
- uv package manager
- API keys (optional - regex search works without any keys)
- Embeddings: VoyageAI (recommended) | OpenAI | Local with Ollama
- LLM (for Code Research): Claude Code CLI or Codex CLI (no API key needed) | Anthropic | OpenAI | Grok (xAI)
Installation
# Install uv if needed
curl -LsSf https://astral.sh/uv/install.sh | sh
# Install ChunkHound
uv tool install chunkhound
Quick Start
- Create
.chunkhound.jsonin project root
{
"embedding": {
"provider": "voyageai",
"api_key": "your-voyageai-key"
},
"llm": {
"provider": "claude-code-cli"
}
}
Note: Use
"codex-cli"instead if you prefer Codex. Both work equally well and require no API key.
- Index your codebase
chunkhound index
- Search changed code in recent commits
# Last N commits
chunkhound search "authentication changes" --last-n 20
# Changes introduced by that commit (diff against its parent; root commits use empty tree)
chunkhound search "database migration" --commit-hash abc1234
# Custom git range
chunkhound search "API changes" --commit-range v2.0..HEAD
# Deep research over recent changes
chunkhound research "what changed in the auth module?" --last-n 50
--vector-sourcecontrols scope:diff(default, changed code only),both(merges diff + DB),db(ignore diff).
For configuration, IDE setup, and advanced usage, see the documentation.
Why ChunkHound?
| Approach | Capability | Scale | Maintenance |
|---|---|---|---|
| Keyword Search | Exact matching | Fast | None |
| Traditional RAG | Semantic search | Scales | Re-index files |
| Knowledge Graphs | Relationship queries | Expensive | Continuous sync |
| ChunkHound | Semantic + Regex + Code Research | Automatic | Incremental + realtime |
Ideal for:
- Large monorepos with cross-team dependencies
- Security-sensitive codebases (local-only, no cloud)
- Multi-language projects needing consistent search
- Offline/air-gapped development environments
License
MIT