Skip to main content
The Plan tab with a proposal waiting: 16 tests ticked, grouped by use case, with Apply Only and Apply and Run in the header

How a Plan Comes About

After the agent has explored your app, it writes a test plan: test cases grouped by use case, each with a title, steps, and a priority set by how much a failure would hurt your customers. A first plan covers each area with the flows that matter there and leaves the long tail for later. The plan arrives as a proposal. The Plan tab, beside the chat, shows the project’s test plan, and changes the agent proposes show there first. Nothing changes on TestSprite until you apply it. You can ask for a plan at any time:
  • Type /plan in the chat box, with notes if you like (“/plan the billing settings page”).
  • Click Plan the Tests on an empty Plan tab.
  • Click Add Tests with AI on Test Cases. The request goes to the agent, and the tests come back as a proposal here.

The Proposal Card

When a proposal is ready, a card appears above the chat box: N changes to the Test Plan. “From this session’s agent. Your tests stay as they are until you apply.”
The proposal card, 16 changes to the Test Plan, above the chat box, beside the Plan tab
With its message, the agent says what it left for later and which expectations it is unsure of. Read that before you apply: an expectation the agent took from what the app does today is a guess until you confirm it.

Reviewing the Changes

The Plan tab shows the whole plan by use case, with the proposal’s changes marked where they are. The counts in the header read like a diff: + new, ~ changed, − removed.
1

Open a changed test

Each test that changes carries a badge and opens to show what changes: its use case, priority, and steps. Tests the proposal leaves alone sit folded under Unchanged.
A proposed test opened on the Plan tab, showing its steps
2

Untick what you don't want

Every change has a tick. Untick one and it is marked Left out: it isn’t applied, and the rest still are.
Two tests unticked and marked Left out; their use case's tick shows part-ticked
3

Edit before you apply

Click Edit Test Plan to change titles, steps, priorities or use cases yourself. The editor has a Form tab and a YAML tab. Your edits stay yours until you save them.
Edit Test Plan on the Plan tab, the Form tab: a test's priority, name, goal and steps as fields
Edit Test Plan on the Plan tab, the YAML tab: the whole plan as one file
4

Apply

Apply and Run saves the plan and starts the run; Apply only saves.
Not happy with the proposal as a whole? Reject it and say what to change (“too many negative cases on login; cover password reset instead”). The agent rewrites it rather than you pruning it row by row.

Running the Tests

A run always waits for you. If you applied without running, or ask the agent to run something later, a card reads Ready to run N tests, lists them, and names the environment. Each test spends credits as it runs.
The Ready to run 16 tests card above the chat box, with Not Now and Run
  • If the project has more than one environment, pick the one to run on from the card.
  • Click Run to start, or Not Now to hold off.
Runs dock under the chat box with their progress and a Watch or View Run button. The agent doesn’t see the results on its own: when a run has failures, click Have the Agent Look (or type /failures) and it judges each one. It sorts every failure into one of three: the app is wrong, the test is wrong, or something got in the way.

Changing the Plan Later

The agent never changes what a test expects just to make it pass. If it can’t tell whether your app or the test is wrong, it asks.

What a Good Test Case Looks Like

The agent writes each case so it can run alone:
  • It creates what it needs and deletes it at the end.
  • It doesn’t rely on another case having run.
  • On production, it avoids steps that could save real data.
Keep these in mind when you edit a case by hand.

Troubleshooting

The project isn’t created yet, or the agent hasn’t explored it. Finish Setup, then click Setup Is Done, Plan the Tests.
The agent may not have reached it. Check the Exploration tab, tell the agent where the feature lives, and ask it to plan tests for it.
Tell the agent what the app really shows (“the button is labelled Save, not Submit”). It fixes the test and says what it changed.
Attach a PRD or spec, or tell the agent your rules (“orders move pending → confirmed → shipped”). It records them in the notes under what counts as correct, and plans against them.
Applying saves the test cases to the project. Ask the agent to remove or revert the ones you don’t want, or delete them on Test Cases.

Where to Go Next

Test Generation & Execution (UI)

What happens when the tests run

Step-by-Step Replay

What you’ll see for each test in a run

Managing Test Cases

Editing and organizing cases after they exist