ChunkHound

Local-first codebase intelligence tool that uses semantic code chunking (cAST algorithm), tree-sitter parsing, and multi-hop semantic search to analyze code repositories. Extracts architecture, patterns, and institutional knowledge from codebases. Supports semantic search, regex search, and LLM-powered code research across 30+ languages.

Evaluated Mar 06, 2026 (0d ago) v4.0.1
Homepage ↗ Repo ↗ Developer Tools code-intelligence mcp semantic-search tree-sitter embeddings codebase-analysis local-first python multi-language
⚙ Agent Friendliness
72
/ 100
Can an agent use this?
🔒 Security
71
/ 100
Is it safe for agents?
⚡ Reliability
62
/ 100
Does it work consistently?

Score Breakdown

⚙ Agent Friendliness

MCP Quality
72
Documentation
74
Error Messages
55
Auth Simplicity
82
Rate Limits
65

🔒 Security

TLS Enforcement
85
Auth Strength
70
Scope Granularity
62
Dep. Hygiene
75
Secret Handling
65

Code chunk search/RAG MCP. Indexes codebase for semantic search. Proprietary code needs self-hosted deployment. No auth for local use.

⚡ Reliability

Uptime/SLA
62
Version Stability
65
Breaking Changes
60
Error Recovery
62
AF Security Reliability

Best When

You work with large or unfamiliar codebases and want AI-powered semantic understanding beyond simple text search, especially with Claude Code or other MCP-compatible editors.

Avoid When

Your codebase is small enough that grep/ripgrep suffices, you cannot run local embedding models and don't want API costs, or you need instant zero-setup search.

Use Cases

  • Semantic natural language search across large codebases (e.g. 'find authentication code')
  • Discovering interconnected code relationships via multi-hop semantic search
  • Understanding unfamiliar codebases through AI-powered architecture analysis
  • Regex-based code pattern matching without requiring API keys
  • Real-time code indexing with file watching for development workflows
  • Cross-team dependency analysis in large monorepos

Not For

  • Simple text search - overkill for basic grep-style needs
  • Projects with fewer than a handful of files - overhead not justified
  • Environments where embedding API costs are a concern and local Ollama is not viable

Interface

REST API
No
GraphQL
No
gRPC
No
MCP Server
Yes
SDK
No
Webhooks
No

Authentication

Methods: api_key
OAuth: No Scopes: No

Optional API keys for embedding providers (VoyageAI, OpenAI) and LLM providers (Anthropic, OpenAI). Can run fully locally with Ollama embeddings and Claude Code CLI or Codex CLI (no API key needed for those).

Pricing

Model: open_source
Free tier: Yes
Requires CC: No

MIT licensed. Embedding and LLM API costs are separate if not using local providers. Fully free with Ollama + CLI-based LLM.

Agent Metadata

Pagination
unknown
Idempotent
Full
Retry Guidance
Not documented

Known Gotchas

  • Requires Python 3.10+ and uv package manager
  • Initial indexing can be slow on large codebases
  • Embedding provider configuration required before first use
  • DuckDB dependency may conflict with other tools
  • Multiple embedding provider options can be confusing for initial setup

Alternatives

Full Evaluation Report

Detailed scoring breakdown, competitive positioning, security analysis, and improvement recommendations for ChunkHound.

$99

Scores are editorial opinions as of 2026-03-06.

5215
Packages Evaluated
26151
Need Evaluation
173
Need Re-evaluation
Community Powered