Cursor-style semantic code search as an MCP plugin for Claude Code.

Instead of grepping 50 files and sending 30,000 tokens to Claude, VecGrep returns the top 8 semantically relevant code chunks (~1,600 tokens) — a ~95% token reduction for codebase queries.

1. Chunk — Parses source files with tree-sitter to extract semantic units (functions, classes, methods)
2. Embed — Encodes each chunk locally using a fine-tuned code search model (384-dim, ~80MB one-time download) via ONNX (~100ms startup) or PyTorch
3. Store — Saves embeddings in LanceDB under ~/.vecgrep/<project_hash>/
4. Search — ANN index (IVF-PQ) for fast approximate search on large codebases

# Install
uv tool install --python 3.12 vecgrep

# Register with Claude Code (once, works for all projects)
claude mcp add --scope user vecgrep -- vecgrep

Then just talk to Claude — it calls VecGrep automatically when you ask about your codebase.