A Model Context Protocol (MCP) server for building and querying document collections with ChromaDB.
- Ingest documents to collections with automatic chunking
uv pip install -e .Run the MCP server:
# Run with persistent storage (required)
uv run main.py --chroma-path /path/to/storage/directoryingest_file(file_path, collection_name, content_type): Read a file, chunk it, and store in ChromaDB- Automatically handles text files with recursive chunking
- Supports binary and image files
- Returns truncated content preview (100 chars), content type, file size, and chunk count
Add this server to your Claude configuration:
{
"mcpServers": {
"doc-builder": {
"command": "uv run /path/to/doc_builder_mcp/main.py --chroma-path /path/to/storage/directory"
}
}
}Run the server in development mode with MCP Inspector:
uv run mcp dev main.py --chroma-path /path/to/dev/storage