Sourcerer
Empowers AI agents to efficiently navigate and search codebases semantically, significantly reducing token consumption.
About
Sourcerer is an MCP server designed to optimize how AI agents interact with codebases. It builds a comprehensive semantic search index of your project, allowing agents to perform conceptual searches and directly access specific functions, classes, or code chunks, rather than processing entire files. This dramatically reduces costly token usage by providing a targeted and intelligent approach to code access. It leverages Tree-sitter for robust code parsing, OpenAI for generating embeddings, and a persistent vector database for efficient storage and retrieval.
Key Features
- Semantic code indexing for conceptual search
- Intelligent code parsing and chunking using Tree-sitter
- Automatic re-indexing of changed files with .gitignore support
- Persistent vector database for storing and querying code embeddings
- MCP tools for targeted code retrieval and index management
- 0 GitHub stars
Use Cases
- Enable AI agents to efficiently find relevant code by concept
- Reduce token costs for large language models interacting with codebases
- Provide targeted code access for AI-powered development tools
Sponsored