DeepContext (Semantic Code Search)

DeepContext provides symbol-aware semantic search for large codebases through AST-based parsing and hybrid vector search, enabling AI assistants to find relevant code chunks by meaning rather than text patterns. Built by Wildcard AI with TypeScript support for Tree-sitter parsing, it combines Jina embeddings for semantic similarity with BM25 full-text search and reranking, while offering both hosted API access and self-hosted deployment with Turbopuffer vector storage. The implementation features incremental indexing with content filtering to exclude test files and generated code, background processing for large codebases, and rate-limited API proxy services, making it valuable for coding agents working with enterprise codebases where traditional grep-based search fails to capture semantic relationships across files.

Connection

• null — =

Source

Repository: https://github.com/wildcard-official/deepcontext-mcp