# How to test a multi-user app when you’re the only developer.

You don’t need a room full of people for every development check. You do need distinct accounts, separate browser sessions and a clear idea of what each person should see. Here is a workflow you can use manually or with Jango’s AI participants.

## 1. Define a two-sided outcome

Choose one interaction rather than asking participants to test everything. For example: Alex invites Maya to a private room; Maya accepts; both see the same room and can exchange a message. Write down the expected result for each account, including what an uninvited account must not see.

## 2. Prepare separate identities

Create development accounts with appropriate roles and separate browser profiles or contexts. Two tabs in the same profile usually share authentication, so they may not represent two people. Prepare the test data and start each account on the relevant screen. In Jango, assign accounts to participants and save their login state once setup is complete.

## 3. Give participants concrete directions

In a manual check, switch between isolated browsers to complete each step. With Jango, give each participant a role and a goal, then join in from your own account. For the invitation scenario, tell Alex whom to invite and tell Maya where to look and what to do after accepting. Set a session budget before starting live AI activity.

## 4. Compare what each account sees

Check visible state rather than trusting a participant’s claim that an action worked. Inspect the destination, member list and message content across browsers. Pause at an unexpected result and record the accounts, steps and expected outcome. A participant getting stuck may expose a confusing interface, or it may be a limitation of the agent.

## 5. Turn confirmed bugs into regression checks

Reproduce the issue with controlled steps. Add a deterministic browser or API test where a precise assertion makes sense. Keep the cast for exploratory runs after future changes, and reset your app’s data when a scenario depends on a clean state. AI exploration, automated regression tests and feedback from real people answer different questions.

[Open Jango](https://usejango.com/#downloads)