Give a browser a job to finish.

Give Molar a website and a task, such as “download the latest invoice” or “check that login works.” The agent works through the site in a real browser and returns the result: a status, the file or fields you asked for, and a recording of every step.

Invoice workflowExample
Browser task01
billing.northstar.test
Find August invoice

August 2026Paid

Page action02
Open matching recordInvoice row selectedinvoice_august_2026
File result03
PDF
invoice_august_2026.pdf Downloaded for the team

Get a result you can use

Every run returns a status (completed, failed or cancelled), any files or fields it collected, and links to its screenshots, video, network log and Playwright trace.

Start from where you work

Run tasks from the dashboard, the molar CLI, the HTTP API or any MCP client.

Sign in with saved logins

Store a test account as a saved login. Molar types it only on your project’s own hosts.

Describe the task and the website

Choose a URL and a clear goal, such as “find the latest invoice and download it.” Molar opens the site, signs in with your saved login, works through the pages and returns what it found.

  • Plain-English goals
  • Saved logins for sign-in
  • Status, files and recordings
After npm install -g @molar/cli and molar login
molar run "download the latest invoice" \
  --url https://billing.example.com --json
Download an invoiceIllustrative example
TaskDownload latest invoice
Open billing portal · done
Find invoice row · done
Save PDF · next
PDFinvoice-downloadRequested filestatus · pending

Call it from your own code

Start the same task over HTTP with an API token from the dashboard, then poll GET /api/v1/agent/tasks/{id} for the result.

HTTP API
curl -X POST https://api.molar.it/api/v1/agent/tasks \
  -H "Authorization: Bearer $MOLAR_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"url":"https://billing.example.com","goal":"download the latest invoice"}'
Read the API reference (OpenAPI)
Run, read the result, fix, run againIllustrative example
Terminal$ molar verify "login works" --url http://localhost:3000 --json
Run outputCheck status and recorded stepstask status · trace referencescreenshots · browser steps inspect and rerun

Where Molar fits

Molar runs both the agent and the browser, then returns a result. You don’t manage browser sessions or write selectors. If you want raw hosted browsers to drive with your own code, a browser infrastructure provider fits better. If you want an agent library inside your own process, use an open-source agent SDK.

Use test accounts

The agent clicks, types and submits forms the way a person would, and it doesn’t pause before a purchase or a delete. It opens pages only on your project’s base URL and the hosts you allow. Point it at test accounts, not at accounts that can move real money.

Frequently asked questions

Have a specific workflow in mind?
Talk to us

What can a browser agent do?

Open pages, sign in with a saved login, fill forms, read information, download files and check that a page shows what you expect.

Can I watch a run?

Yes. While a hosted run is working, its live view shows the browser in real time, read-only. Afterwards, the video, screenshots and trace are kept for 5 days.

Where can the agent go?

Only to pages on your project’s base URL and the hosts you allow. Hosted runs can’t reach private networks.

What does it cost?

Molar is free during the beta, with up to 25 hosted runs a day per organization.

Give a browser a job.

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

Free during the beta. No card.