How agents are driven
RigorRun gives your agent one task at a time and a URL to work through. Your agent connects to that URL, works however it normally works, and says when it is done.
What an agent receives
Section titled “What an agent receives”The public half of a case: the task, in the words you wrote when you created the project, plus the tools it may use. Your agent is given exactly those words, so write them for a reader who knows nothing.
What it never receives
Section titled “What it never receives”The private half: the checks and the reference plan. These are stripped before a case leaves RigorRun, so there is nothing to read the answer from. The split is enforced by the type, and there is a test asserting a public case view carries no assertions.
Three ways to connect
Section titled “Three ways to connect”| When to use it | |
|---|---|
| HTTP | Your agent listens on an address RigorRun can reach. |
| Process | Your agent is a command on this machine; RigorRun starts it. |
| Driven | RigorRun cannot start or reach it — your loop pulls work instead. |
What is scored
Section titled “What is scored”What changed in your system. Your agent’s own report of what it did is displayed on every case, labelled not scored, and never affects a verdict.
That is deliberate and it is the point of the product: an agent that reports success and an agent that achieved it are indistinguishable from the transcript.