Running an agent

The app lists every agent the harness has registered. Each agent has three tabs:

• Run: a form drawn from the agent's input, with example inputs. Submitting it runs the agent through the conductor. The page shows the stages, the outcome, the result and the evidence it rests on.
• About: what the agent takes and returns, how it grounds its results, and whether your deployment's policy profile permits it. Anything an agent has not declared is shown as "Not declared".
• History: the stored runs. Each reopens in the same view.

Parts of the app that are not built yet are marked "Not built yet", with a link to the issue.