Molar.
Get started

AI agents that
use the web
for you.

Tell Molar what to do: download an invoice, fill in a form, collect data, or test your app. Watch the browser work and review the result.

Start automating See it work
A URL. A goal. An outcome you can inspect.
Molar / Browser workspaceIllustrative demo
Your goal

Find the invoice and download a copy.

demo.acme.test/billingTest environment
Billing is openObserved
acme.finance.testSimulated
1 Invoice list2 Invoice3 Download
Invoices
INV-1042Aug 24, 2026$240.00
INV-1039Jul 24, 2026$240.00
Authenticated billing page visible.
Open the billing page.
Step 1 of 3
Isolated test dataSample walkthrough · No live run
From a single task
to a reusable workflow.Here to test your app?
Meet Molar for QA
Work from the tools
you already use.
CLIMCPREST APIPlaywrightYour coding agent
One browser execution layer

The task is yours.
The browser work is Molar’s.

Choose what you need done. Molar works through the page and brings back a result you can inspect.

Complete the work
behind the login.

Download an invoice, update a record, or fill in a form. Give Molar the task, the website, and the access it needs.

Meet browser agents
Browser taskIllustrative workflow
billing.acme.test
Browser taskFind August invoice

Aug 2026 Paid

Documentinvoice_august_2026.pdf
Downloaded

Turn a page
into useful data.

Read the names, prices, or records you need and return them in a structure your software can use.

Explore data extraction
Structured extractionIllustrative workflow
Plans demo.acme.test/plans

Starter$9monthly

Pro$24monthly

Business$79monthly

3 visible records
extract
{
  "plan": "Pro",
  "amount": 24,
  "currency": "USD",
  "interval": "month"
}
Structured result · source URL retained

See what happened.
Know what to do next.

Watch the browser, follow the steps, and inspect the screenshots and events that explain the outcome.

Explore live view & evidence
Live evidenceIllustrative workflow
checkout.acme.test / confirmation
Order confirmedReceipt sent to test inboxpayment_intent.succeeded
Run trace 00:08—00:10

Open checkoutScreenshot captured200

Confirm paymentPOST /webhooks/payment200

Receive receiptTest inbox matchedready

Network and console evidence attached

A goal goes in.
Evidence comes back.

Give a human or an agent the same foundations: an authorized target, an observable browser, and a result to inspect.

01

Choose the job.

Give Molar a URL and a task. Set the scope, the credentials it can use, and the outcome you expect.

molar run "download the latest invoice"
02

Let the browser work.

Molar observes the page, takes an action, checks what changed, and uses that state to choose its next step.

Observe → act → verify
03

Use the result.

Get a structured outcome and artifact references. Inspect the work, reuse a flow, or hand the evidence to an agent.

Result + screenshots + trace

A browser action
is only half the story.

What changed? Did the page reach the expected state? Where did the journey stop?

Molar pairs browser execution with verification and trace evidence. Your next decision starts from the observed outcome.

Follow the evidence
Checkout / evidenceIllustrative trace
Expected outcome
Order confirmedReceipt sent to test inbox
Observed outcome
Order pendingPayment callback not applied
00:08Payment confirmed200
00:09POST /webhooks/payment500
00:10Order status assertionFailed
ScreenshotNetworkConsoleSteps

Test what each user
can actually do.

A guest shouldn’t see private projects. A member shouldn’t change billing. Plan checks across roles, devices, and account states with Molar Swarm.

Preview Availability is confirmed for your beta setup.

Agent coverageIllustrative workflow
One workspace. Different permissions.Example test plan
GuestPublic access
acme.test
Browse plans
Starter$12 / month
Pro$24 / month
Private projects stay hidden
MemberProject access
acme.test
Acme workspace
Website launchCan edit
API gatewayCan edit
Billing settings stay restricted
AdminWorkspace access
acme.test
People & permissions
PRPriya RaoOwner
SCSam ChenMember
Role changes apply correctly
Three example roles · actual coverage depends on your configured test plan.

Keep the checks you plan separate from the results you’ve observed. See which combinations need another look.

Explore Swarm

A browser for your
next agent workflow.

Start a task, extract structured data, verify a claim, or inspect a trace. CLI and MCP interfaces let your tools work from the same browser evidence.

CLI MCP API
Build with Molar
Terminal
molar run "find the latest invoice" \
  --url https://staging.example.com \
  --json
Structured results Inspectable evidence Scoped access

Install and configure Molar first. Read the quickstart

One journey.
The tools to follow it through.

Explore with Cartographer. Control dependencies with Clones. Check releases with Guard. Understand failures with Trace.

Application mapIllustrative flow
Your application
Guest
Homepage
Browse plans
Create account
Member
Sign in
Workspace
Invite teammate
Observed pathNext branch
PagesActionsOutcomesEvidence

Discover meaningful paths through your application.

Explore Cartographer
And when something breaks? Mender turns failure evidence into a proposed repair.Meet Mender Preview

One platform.
A deeper kind of QA.

Molar for QA combines browser execution with stateful test services, release checks, and debugging. Put a complete user journey under test, including the dependencies behind it.

Explore Molar for QA
Connected test servicesIllustrative setup
Your applicationOne isolated test environment
SStripePayment state
@EmailVerification & receipts
TTwilioSMS & OTP
AAuthSessions & identities
S3StorageFiles & uploads
SnapshotAdvance timeRestore

A few things
you might be wondering.

Have a specific workflow in mind?
Let’s talk it through

What can I do with the Molar platform?

Run authorized browser tasks, explore application flows, extract structured data, verify page outcomes, and inspect the resulting evidence. The QA product uses these same browser foundations.

Who is the platform for?

Developers building browser workflows, teams automating repetitive web tasks, and AI agents that need to use and verify an application.

Can I use natural language and code?

Use plain-language goals through the dashboard, CLI, or MCP, and use Playwright-oriented workflows when you want editable browser code. Availability depends on your deployment and the current beta coverage.

Does a task run without boundaries?

You define the target, authorization, credentials, and goal. Unsupported or sensitive steps can require input. Failed, blocked, and completed are distinct outcomes.

How do I connect my coding agent?

Use the Molar CLI or an MCP-compatible client. Start with the developer documentation to configure the endpoint, credentials, and scopes for your environment.

What should the web
do for you?

Start with a concrete task. Give your team and agents a browser they can work with.

Get started Talk to the team