Use case

Many agents, using your product like real users

Real products are used by different people at the same time: an admin, a collaborator, someone who should not have access. Composal sends agents in as each of them, with their own accounts, and watches how they affect one another.

Read the docs
Admin
Shares a project with a teammateDone
Collaborator
Opens the shared project and edits itIn progress
Visitor
Tries to open the same link, signed outBlocked, as intended

Where the time goes today

Tests log in as one user

A single test account cannot catch what happens when two people work on the same thing at once.

Permissions are checked by hand, or not at all

Who can see and change what is checked rarely, and almost never as the person you are worried about.

Test accounts are clean, production is not

Real users have years of history, odd data, and feature flags. A fresh account hides the bugs that live there.

1

Give each agent a role

Personas have their own sign-in, goals and permissions: admin, collaborator, visitor. Add as many as your product has.

AdminSets up the workspace and invites people
CollaboratorEdits shared work and leaves comments
VisitorTries to reach what they should not
2

Let them act on the same product

In one journey, one persona acts and another has to see the result. Composal checks each side, in order, with each agent’s evidence.

AdminShares a project
CollaboratorSees it in their list
3

Use accounts that look like production

Persistent accounts with realistic data, feature flags and edge cases, so what the agents find is what your users would.

Persona account
Own sign-inRealistic dataFeature flagsPast activity
Accounts that look like production, not an empty test org.

Bugs between users

The sharing, notification and permission bugs a single test account cannot see.

Permissions that hold

Every persona is checked for what it can and cannot do.

Production-shaped findings

Realistic accounts mean realistic bugs, not test-data noise.