Why Harness AI Uses a Knowledge Graph, Not Raw APIs

Harness employs a schema-driven Knowledge Graph to enhance the performance and reliability of AI agents operating across its multi-module platform, which includes CI/CD, DevSecOps, and FinOps. This approach offers a more efficient alternative to the Model Context Protocol (MCP) that relies on raw API calls, which can lead to high token costs, latency, and errors. By utilizing the Harness Query Language (HQL) to access structured platform data, the Knowledge Graph ensures deterministic and low-latency answers, reducing token usage by up to 25 times compared to MCP. The Knowledge Graph stores comprehensive metadata for fields, explicitly declares cross-module relationships, and uses semantic annotations to efficiently route queries, thus minimizing the need for the AI to infer connections or interpret data incorrectly. This structured approach not only enhances the AI's ability to retrieve and analyze data but also supports a tiered data ownership model that prioritizes using the Knowledge Graph for maximum reliability and cost-effectiveness.