# graph-tool-call

> Graph-structured tool retrieval for LLM agents.

## Documentation

- Home: https://sonaiengine.github.io/graph-tool-call/
- Overview: https://sonaiengine.github.io/graph-tool-call/docs/
- Quickstart: https://sonaiengine.github.io/graph-tool-call/docs/getting-started/quickstart/
- Mental Model: https://sonaiengine.github.io/graph-tool-call/docs/getting-started/mental-model/
- OpenAPI Search-To-Plan Tutorial: https://sonaiengine.github.io/graph-tool-call/docs/tutorials/openapi-search-to-plan/
- API Cheat Sheet: https://sonaiengine.github.io/graph-tool-call/docs/reference/api-cheat-sheet/
- Tool Graph Search: https://sonaiengine.github.io/graph-tool-call/docs/search/tool-graph-search/
- OpenAPI Ingestion: https://sonaiengine.github.io/graph-tool-call/docs/build/openapi-ingestion/
- IO Contracts: https://sonaiengine.github.io/graph-tool-call/docs/build/io-contracts/
- Target Selection: https://sonaiengine.github.io/graph-tool-call/docs/search/target-selection/
- Plan Synthesis: https://sonaiengine.github.io/graph-tool-call/docs/plan/plan-synthesis/
- Trace Learning: https://sonaiengine.github.io/graph-tool-call/docs/concepts/trace-learning/
- Quality Lab: https://sonaiengine.github.io/graph-tool-call/docs/validation/quality-lab/
- XGEN Integration: https://sonaiengine.github.io/graph-tool-call/docs/guides/xgen-integration/
- Public API: https://sonaiengine.github.io/graph-tool-call/docs/reference/public-api/
- Benchmarks: https://sonaiengine.github.io/graph-tool-call/docs/validation/benchmarks/

## Package

- PyPI: https://pypi.org/project/graph-tool-call/
- Source: https://github.com/SonAIengine/graph-tool-call

## Summary

graph-tool-call builds tool graphs from OpenAPI, MCP, and Python function
sources. It uses keyword retrieval, graph expansion, semantic metadata, contract
evidence, target selection, plan diagnostics, and scrubbed trace learning to
help LLM agents choose and use tools from large catalogs.