Sourcerer

Indexes codebases using Tree-sitter parsing and vector embeddings to enable semantic code search through natural language queries

st3v3nmw 118
Claude CodeClaude DesktopGeneric
View source ↗

This MCP server provides semantic code search and analysis for Go codebases using Tree-sitter parsing and vector embeddings, built by Stephen Mwangi with chromem-go for vector storage and real-time file watching capabilities. The implementation indexes Go source files into semantic chunks (functions, methods, types) with automatic staleness detection and background re-indexing, offering two core tools: semantic search that finds relevant code sections using natural language queries, and source code retrieval that returns formatted code chunks with line numbers and summaries. Features include Tree-sitter-based AST parsing for accurate code structure extraction, file system watching with debounced updates for active development workflows, and persistent vector storage that maintains search indexes across sessions, making it valuable for developers exploring unfamiliar codebases, conducting code reviews, or building AI-assisted development tools that need contextual code understanding.

Source

Repository: https://github.com/st3v3nmw/sourcerer-mcp

Maintain Sourcerer?

Let people know it's listed here — add the badge (live metrics, light/dark aware) or a plain link to your README or docs.

[Sourcerer on getagentictools](https://getagentictools.com/mcp/st3v3nmw-sourcerer-mcp?ref=badge)