Let the agent verify the app it just changed.

After an agent changes a checkout or login flow, it needs to use the app. Molar gives it CLI and MCP tools to run a concrete browser task, read the status and captured output, fix the code, and try again.

Get started free Read the docs

A check from your terminalExample

$ molar verify "the login form is visible" \
--url http://localhost:3000 --json

{
  "ok": true,
  "pass": false,
  "failure_taxonomy": "element_not_found",
  "rationale": "The page shows a Sign in button but no password field.",
  "hint": "Check that /login renders the form for signed-out visitors.",
  "failure": { "code": "element_not_found", "retryable": false },
  "execution_mode": "local",
  "artifact_manifest_url": "https://api.molar.it/v1/runs/…/artifacts/manifest"
}

One command after code

Run molar verify "the login form is visible" --url http://localhost:3000 --json and pass the status to the next agent turn.

Connect through MCP

Use https://api.molar.it/mcp or the local stdio server through molar mcp serve, with the access your target requires.

Use test accounts

Runs open pages only on your project’s hosts. The agent doesn’t pause before a purchase or a delete, so give it test accounts and Stripe test mode.

Get output your agent can read

The status, run ID, Trace link, and captured screenshot give an agent enough context to diagnose and rerun. --json keeps the result machine-readable.

JSON status
Run polling
Screenshots and Trace links

Run, read the result, fix, run againIllustrative example

Terminal$ molar verify "login works" --url http://localhost:3000 --json

Run outputCheck status and recorded stepstask status · trace referencescreenshots · browser steps inspect and rerun

Check after every code change

Change the login form, run the local check, read the pass or failure, inspect the captured screenshot, and rerun after the fix. Hosted runs use the target and credentials you choose.

After npm install -g @molar/cli and molar login

molar verify "the login form is visible" \
  --url http://localhost:3000 --json

Know what the result meansRun status

Completed: The requested result was observed.
Failed: The expected page state was not reached.
Blocked: Access, a required input, or a workflow limit prevented progress.

Review the recorded steps and screenshots to see where the task stopped.

Ask for a result you can see

Ask for “show the invoice number,” “reach the signed-in dashboard,” or “confirm the order says paid.” Molar returns the run identity and available output for the next turn.

Visible business outcomes
Trace links
Structured failure hints

Frequently asked questions

Have a specific workflow in mind?
Talk to us

Can an agent click a payment button automatically?

Yes. The agent doesn’t pause before a payment or a delete in the browser. Use test accounts, and Stripe test mode or the Stripe clone.

Can an agent use a real user's login?

Use a saved test account or credential reference for the site. Supported email and SMS connections provide verification codes without using a customer’s personal inbox.

What happens when verification fails?

The CLI can emit a structured failure with a hint plus screenshot and Trace links so the coding agent can inspect the cause and rerun.

Let your agent test its changes.

Create a free workspace and describe the first flow to test. Or check a public page first, no account needed.

Get started free Check a public page

Free during the beta. No card.