Codeweaver

Codeweaver is a local MCP server that gives an AI assistant semantic, AST-aware code search across a project, covering 166 languages.

What you can do

Search a codebase semantically rather than by exact text, using hybrid retrieval over code structure and meaning. Work across 166 languages with AST-aware indexing that understands code structure. Configure embedding, reranking, sparse-embedding and vector-store providers to tune how results are retrieved. Point it at a specific project path or name so the assistant searches the right codebase.

When to use it

Useful when an assistant needs to find relevant code in a large or unfamiliar repository where keyword search falls short, and you want retrieval quality improved with embeddings and reranking.

Requirements

Runs locally as a stdio process, launched with uvx from the code-weaver PyPI package or from the Docker image. Needs a project path and a chosen embedding, reranking and vector-store provider; API keys for those providers are supplied through environment variables such as CODEWEAVER_EMBEDDING_API_KEY, CODEWEAVER_RERANKING_API_KEY and CODEWEAVER_VECTOR_STORE_API_KEY, or through cloud provider credentials for AWS Bedrock, Azure OpenAI, Azure Cohere or Vercel AI Gateway.