AI Harness / ecosystem explorer

The AI Harness ecosystem

Twelve components around one model. Select any component for its definition, assurance risk, governing control, accountability and evidence. Trace the operating flow or the evidence flow one relationship at a time. This is the final step of the AI Harness lab, presented on its own so it can be expanded and explored.

JavaScript draws the interactive ecosystem. Without it, the AI Harness components are: inputs and external content; orchestrator / agent loop; RAG subsystem; memory and state; context assembly; input / output guardrails; policy / authorization gate; credential broker; extensions / MCP / skills; tool dispatcher; sandbox / downstream systems; observability and audit; and the generative model at the core. Read the assurance domains or the glossary for the definitions.