Your agents build it.Ours break it.

Composal browser agents test real user flows and edge cases in your product. See what works, what breaks, and whether your changes are ready to ship.

Release confidently

Know your critical user flows work before you ship

Ship faster

Cut the manual testing work before every release

Reduce bugs

Catch issues before they reach your users

1

Map your product

Give Composal your URL. Connect your repo, specs, and other sources of product intent - Notion, Posthog, Granola

Discovery
2

Launch an agent swarm

Deploy a swarm of agents to explore your product, test key flows, and uncover edge cases in parallel

Deep Scan

Messages

Is the 5:30 still open?

Yes, two seats left.

I'll take both.

Message

Signups

+18%

1,284

M
T
W
T
F
S

October

Thu 2 · 5:30pm

MTWTFSS
12345678910111213141516171819202122232425262728293031

Meet

Belém dockHarbor kayak · 90 min

Payment

Card

•••• •••• •••• 4242

Expiry

10 / 28

CVC

•••

Pay €36

Queue

New

Confirm kayak
Send waiver

Doing

Charge card
Check weather

Done

Welcome email

Guest waiver

Harbor kayak · 2 Oct

Vests are at Belém dock. Arrive 15 minutes early.

The session runs 90 minutes and stays inside the harbor.

Signed · Ana Costa

Preferences

Email reminders
SMS updates
Share location
3

Inspect the outcomes

See what worked and what broke: video, logs and the steps to reproduce every failure

Scenarios

Harbor kayak booking

Failed

checkout.lumen.dev

  1. Opened Lisbon0:02
  2. Selected Harbor kayak0:06
  3. Chose Thu 2 Oct, 2 guests0:09
  4. Couldn’t complete booking0:11
4

Send failures back to your agent

Pass the reproduction steps to your coding agent so it can fix the issue and learn from it

CLI