Skip to main content
Some sites block automation, require login, show CAPTCHA, or need human review. A good browser-agent workflow reports that state clearly.

​
What the agent should do

  • Return the blocker status.
  • Include the visible page evidence.
  • Stop before out-of-scope actions.
  • Ask for confirmation when configured.
  • Avoid inventing missing data.

​
CLI review path

web agent \
  --debug-url \
  --approval ask \
  --url https://www.nhtsa.gov/recalls \
  --allow-domain nhtsa.gov \
  "Search recalls for a sample 2020 Toyota Camry query. Do not enter a real VIN or personal data."

​
SDK handling

const result = await agent.run({
  startUrl: "https://www.nhtsa.gov/recalls",
  goal: "Search recalls for a sample 2020 Toyota Camry query. Do not enter a real VIN.",
});

if (result.status === "blocked") {
  await queueReview({
    browserId: result.browserId,
    summary: result.text,
    steps: result.steps,
  });
}
Do not document CAPTCHA solving as the happy path. If a run needs human or provider-specific resolution, make that review step explicit.