Skip to lesson
Exit
How Coding Agents Work (and Their Limits)1 / 3

1 min lesson

The loop, step by step

Use "An agent run is not one model call" to say what you would do next.

Step 1 of 3

The loop, step by stepprompt → applied edit

An agent run is not one model call. It is a loop that keeps calling the model, feeding back what each action produced, until the task is done or a stop condition trips.

Learn more

Full explanation

The Agent Loop

THE AGENT LOOP

Interactive diagram. Step through it with the Next and Previous controls below, or Tab to a region to read its detail.

diagram: flow

The whiteboard artifact: assemble, decide, act, observe, revise - the last step loops back until the task is done.

The harness is the product, not the model

Anyone can call a high-capability model's API. What makes Cursor an agent - and what makes one agent better than another on the same model - is the harness: how it indexes a million-line repo, what it retrieves into a finite context, how it defines tools the model can call reliably and how it decides when to stop. When a developer asks 'which model does Cursor use,' the more interesting answer is 'here's what we do around the model.'

Ground that claim in what Cursor's harness actually does. It is not a thin desktop app piping your prompt straight to a model provider. There is deep proprietary cloud infrastructure - the Cursor harnessCursor's hosted layer around each model (context selection, caching, retries) that makes the same model run better and cheaper than calling it directly. Press Enter for the full definition. - doing optimization, context enrichment, caching and compaction around the model, tuned per frontier model.