Cursor: The agentic editor that made agent state visible
Cursor turned coding into an agent task — and solved the trust problem with a persistent agent status, auditable tool calls and an explicit approval gate.
An independent reading of Cursor's interface, grounded in Cursor's official documentation. Quotes are verbatim from the linked official pages. The product UI may differ from what is described.
Agent status, always in view
- The agent's state — thinking, running, waiting — is a persistent quiet indicator, never a blocking modal.
- State changes announce themselves in place; the diff updates live as files change.
“Watch edits appear in the diff view as they happen.”
Source: Cursor docs — Agent mode ↗
Searching 12 sources for recent AI model releases…
Tool calls you can audit
- Every file read, write and shell command lists with its outcome — the agent's work is fully auditable.
- Expand to see the diff or command output; collapse to keep the flow.
“Cursor includes tools that help agents write code: reading files, editing files, running terminal commands, searching the web, and more.”
Source: Cursor docs — Agent Security ↗
The approval gate
- Irreversible actions pause the agent until a human approves — the exact command is shown before it runs.
- Approve and deny are explicit and one click away; the default is to ask, not to act.
“Run Modes control how the Cursor agent runs tool calls, and when Cursor interrupts you for approval.”
Source: Cursor docs — Run Modes ↗
Run database migration
Deploy Agent is requesting approval
Runs an irreversible migration on the production database. 4 tables, ~2 min estimated downtime.
npm run migrate -- --env=production --confirm
Diff-first communication
- The agent communicates by editing, and the diff is the message — progress lives in the file tree, not a chat log.
- Chat scrolls out of the way; the workspace is the primary surface.
“Agent's edits are applied as it works. Review them in the diff view and reject anything you don't want.”
Source: Cursor docs — Agent mode ↗
Design takeaways
- Keep agent state visible but quiet; announce changes in place, never in a modal.
- Auditability beats automation theater: show every action and its outcome.
- Gate irreversible actions behind explicit approval, with the exact command shown.
- Communicate by editing — make the diff the conversation, the workspace the surface.