graphrag-rs

Turn any folder of documents into a knowledge graph — then query it from Claude, Cursor, or any MCP-compatible AI assistant.

graphrag-rs is a single Rust binary that:

1. Reads your docs (.md, .txt, .csv, .json)
2. Extracts entities and relationships via LLM
3. Serves the knowledge graph over MCP (Model Context Protocol)

Powered by Samyama, a high-performance embedded graph database.

curl -sSL https://raw.githubusercontent.com/samyama-ai/graphrag-rs/main/install.sh | sh

Or build from source:

# graphrag-rs depends on samyama-graph via relative path — clone both side by side
mkdir -p graphrag-workspace && cd graphrag-workspace
git clone https://github.com/samyama-ai/samyama-graph.git
git clone https://github.com/samyama-ai/graphrag-rs.git
cd graphrag-rs
cargo build --release
cp target/release/graphrag-rs ~/.local/bin/

export OPENAI_API_KEY="sk-..."
graphrag-rs ingest ./my-docs/

That's it. graphrag-rs walks the folder, chunks the text, extracts entities and relationships via GPT-4o-mini, and saves the graph locally.

graphrag-rs serve

This starts an MCP server over stdio. Add it to your assistant's config:

Claude Desktop (~/Library/Application Support/Claude/claude_desktop_config.json):

{
  "mcpServers": {
    "graphrag": {
      "command": "graphrag-rs",
      "args": ["serve"]
    }
  }
}

Claude Desktop (~/Library/Application Support/Claude/claude_desktop_config.json on macOS):

{
  "mcpServers": {
    "graphrag": {
      "command": "graphrag-rs",
      "args": ["--data-dir", "/path/to/your/data", "serve"]
    }
  }
}

> Restart Claude Desktop after editing. The graphrag tools will appear under the hammer icon.

Claude Code (.mcp.json in your project dire