Local semantic code search for Claude Code. Index your codebase once, search it with natural language.

# npm
npm install -g claude-codemem

# homebrew (macOS)
brew tap MadAppGang/claude-mem && brew install --cask claudemem

# or just curl it
curl -fsSL https://raw.githubusercontent.com/MadAppGang/claudemem/main/install.sh | bash

Claude Code's built-in search (grep/glob) works fine for exact matches. But when you're trying to find "where do we handle auth tokens" or "error retry logic" — good luck.

claudemem fixes that. It chunks your code using tree-sitter (so it actually understands functions/classes, not just lines), generates embeddings via OpenRouter, and stores everything locally in LanceDB.

The search combines keyword matching with vector similarity. Works surprisingly well for finding stuff you kinda-sorta remember but can't grep for.

# first time setup
claudemem init

# index your project
claudemem index

# search
claudemem search "authentication flow"
claudemem search "where do we validate user input"

That's it. Changed some files? Just search again — it auto-reindexes modified files before searching.

Run your own benchmark with claudemem benchmark. Here are results on real code search tasks:

Summary:

• 🏆 Best Quality: voyage-code-3 (175% NDCG)
• ⚡ Fastest: voyage-3.5 (1.2s)
• 💰 Cheapest: all-minilm-l6-v2 (local, free)

claudemem supports three embedding providers:

claudemem init  # select "OpenRouter"
# requires API key from https://openrouter.ai/keys
# ~$0.01 per 1M tokens

# install Ollama first: https://ollama.ai
ollama pull nomic-embed-text

claudemem init  # select "Ollama"

Recommended Ollama models:

• nomic-embed-text — best quality, 768d, 274MB
• mxbai-embed-large — large context, 1024d, 670MB
• all-minilm — fastest, 384d, 46MB

claudemem init  # select "Custom endpoint"
# expects OpenAI-compatible /embeddings endpoint

View available models:

claudemem --models           # OpenRouter models
claudemem --models --ollama  # Ollama models

Run it as an MCP server:

claudemem --mcp

Then Claude Code can use these tools:

• search_code — semantic search (auto-indexes changes)
• index_codebase — manual full reindex
• get_status — check what's indexed
• clear_index — start fresh

This repo also contains an experimental VS Code inline completion extension that talks to a persistent claudemem autocomplete server.

• Autocomplete server: claudemem --autocomplete-server --project .
• VS Code extension source: extensions/vscode-claudemem-autocomplete/

1. Parses code with tree-sitter — extracts functions, classes, methods as chunks (not dumb line splits)
2. Generates embeddings via OpenRouter (default: voyage-3.5-lite, best value)
3. Stores locally in LanceDB — everything stays in .claudemem/ in your project
4. Hybrid search — BM25 for exact matches + vector similarity for semantic. Combines both.

TypeScript, JavaScript, Python, Go, Rust, C, C++, Java.

If your language isn't here, it falls back to line-based chunking. Works, but not as clean.

claudemem init              # setup wizard
claudemem index [path]      # index codebase
claudemem search <query>    # search (auto-reindexes changed files)
claudemem status            # what's indexed
claudemem clear             # nuke the index
claudemem models            # list embedding models
claudemem benchmark         # benchmark embedding models
claudemem --mcp             # run as MCP server

Search flags:

-n, --limit <n>       # max results (default: 10)
-l, --language <lang> # filter by language
-y, --yes             # auto-create index without asking
--no-reindex          # skip auto-reindex

Env vars:

• OPENROUTER_API_KEY — for OpenRouter provider
• CLAUDEMEM_MODEL — override embedding model

Files:

• ~/.claudemem/config.json — global config (provider, model, endpoints)
• .claudemem/ — project index (add to .gitignore)

• First index takes a minute on large codebases
• Ollama is slower than cloud (runs locally, no batching)
• Embedding quality depends on the model you pick
• Not magic — sometimes grep is still faster for exact strings

MIT

GitHub · npm · OpenRouter