Runs

A run is one request against a Spinup Agent. You send input, Spinup picks an environment and a harness for that request, and the harness produces output.

Runs are the current billable unit.

What a run records

Every run keeps a durable record of what actually happened:

• the selected harness and the requested harness (when an override was used)
• the model that was called
• the runtime instance that served the request
• the projected state the harness saw at the start of the run
• the output, including stdout, stderr, and any warnings
• evidence of which capabilities were available and which proposed changes were accepted

That record is what makes runs comparable across harnesses and reproducible for support.

What a run does not change

A run does not change the Spinup Agent on its own. Memory, capabilities, secrets, and harness settings only change when the agent's saved configuration is updated, or when a harness submits a proposed change and Spinup accepts it. The run's output is output, not state.

Per-request overrides such as a different harness or a different model apply only to that one run.

Where to go next

• See Spinup Agents for the object a run targets.
• See Environments for the runtime capacity that serves a run.
• See Memory for how a harness asks Spinup to keep something from a run.
• Send a request and read the response shape in Status and exec.