Sourcerer MCP
Semantic code search & navigation for AI agents — jump to the exact functions and chunks instead of reading whole files.
Add to your client
Copy the config for your MCP client and paste it into its config file.
go install github.com/st3v3nmw/sourcerer-mcp/cmd/sourcerer@latestPaste into ~/Library/Application Support/Claude/claude_desktop_config.json
{
"mcpServers": {
"sourcerer-mcp": {
"command": "sourcerer",
"args": [],
"env": {
"OPENAI_API_KEY": "your-openai-api-key",
"SOURCERER_WORKSPACE_ROOT": "/path/to/your/project"
}
}
}
}Step-by-step guides: Add to Claude Desktop · Add to Cursor · Add to Windsurf
Before you start
- OpenAI API key (for generating embeddings; local embedding support is planned)
- Git — the workspace must be a git repository (respects .gitignore files)
- Add .sourcerer/ to .gitignore (stores the embedded vector database)
- Install the sourcerer binary via Go (go install) or Homebrew (st3v3nmw/tap)
About Sourcerer MCP
Sourcerer MCP gives AI agents semantic search and navigation over a codebase so they can find and retrieve only the relevant functions, classes, and code chunks instead of reading whole files — dramatically reducing token usage. It uses Tree-sitter to parse files into ASTs and extract chunks with stable IDs, watches the filesystem (fsnotify) and respects .gitignore for automatic re-indexing, and stores OpenAI-generated embeddings in a persistent chromem-go vector database under .sourcerer/db/.
Tools & capabilities (5)
semantic_searchFind relevant code using semantic search.
get_chunk_codeRetrieve specific chunks by ID.
find_similar_chunksFind similar chunks.
index_workspaceManually trigger re-indexing of the workspace.
get_index_statusCheck indexing progress.
When to use it
- Let AI agents find relevant code conceptually without reading entire files, reducing token usage and cognitive load
- Jump directly to specific functions, classes, methods, or types by semantic relevance
- Discover code similar to a given chunk across the codebase
- Keep a semantic index automatically up to date as files change in a git repository
Security notes
Requires an OpenAI API key (passed via the OPENAI_API_KEY environment variable) to generate embeddings; code chunks are sent to OpenAI's API for embedding. Must be run inside a git repository and respects .gitignore. Add the .sourcerer/ directory (which stores the embedded vector database) to .gitignore.
Sourcerer MCP FAQ
Do I need an OpenAI API key?
Yes. An OpenAI API key is required to generate embeddings. Local embedding support is planned.
Which languages are supported?
Go, JavaScript, Markdown, Python, and TypeScript are supported. C, C++, Java, Ruby, Rust, and others are planned.
How do I install it?
Install the binary with `go install github.com/st3v3nmw/sourcerer-mcp/cmd/sourcerer@latest`, or via Homebrew: `brew tap st3v3nmw/tap` then `brew install st3v3nmw/tap/sourcerer`.
How do I add it to Claude Code?
Run: `claude mcp add sourcerer -e OPENAI_API_KEY=your-openai-api-key -e SOURCERER_WORKSPACE_ROOT=$(pwd) -- sourcerer`.
Where is the index stored?
In the .sourcerer/db/ directory using chromem-go for persistent vector storage. Add .sourcerer/ to your .gitignore.
Alternatives to Sourcerer MCP
Compare all alternatives →
Microsoft's official browser-automation MCP using Playwright's accessibility tree (no vision model).
Up-to-date, version-specific library documentation injected into your coding agent.
LSP-powered coding agent toolkit: semantic symbol search, references and structural edits.
Compare Sourcerer MCP with: