Harness
Also known as: agent harness · scaffolding
A harness is everything around the model that turns it into an agent: the loop it runs in, the tools it can call, its memory, its rules.
- The model decides; the harness executes and feeds results back.
- Includes: system prompt, tool registry, loop, memory store, guardrails.
- Two harnesses around the same model can behave very differently.
The key idea: the model is the *brain*, the harness is the *body*. A lot of an agent's behavior — good or bad — lives in the harness, not the weights.