What it does
Embgrep indexes code and documentation files into local embeddings so they can be searched by meaning rather than exact keywords. It uses fastembed on ONNX Runtime with the BGE-small-en-v1.5 model, producing 384-dimensional vectors stored in a single SQLite file. Code is split by function or class, documents by heading or paragraph, and config files into fixed-size chunks. Supported types include .py, .js, .ts, .java, .go, .rs, .md, .txt, .yaml, .json and .toml.
It suits an LLM agent that needs to locate relevant files in a codebase or documentation set by intent rather than by exact text. The MCP server exposes four tools: index_directory, semantic_search, index_status and update_index. Re-indexing is incremental, comparing SHA-256 hashes so only changed files are processed. A Python API and a CLI are also available.
Running it means adding the server to an MCP client configuration such as Claude Desktop, either directly or through uvx. Search returns the top 5 results by default, and the database path and model are configurable. Indexing and queries run locally.
Safety
Our team has not reviewed this resource. Only automated rules checked it; resources with high-risk signals are not published. These are signals, not a guarantee that installing it is safe. Try it in a sandbox project first.
- Runs code or changes an assistant's behaviour
Sources
- Official MCP Registry first seen 13 Nov 2026 · last seen 13 Nov 2026 · its page on the source
Embgrep
Local semantic search — embedding-powered grep for files, zero external services.
v0.1.0As shown on registry.modelcontextprotocol.io when we last read it