Tester: AI testing for your Fortnite map, backed by evidence
The Tester is your team’s QA specialist. For AI testing of your Fortnite map, it proves or disproves a specific behavior with reproducible evidence, and tells you plainly what was checked and what was not.
AI testing for your Fortnite map
The Tester reproduces bugs, designs and runs available checks, writes scoped test cases and reports regressions. It also reviews device wiring, gameplay math and playtest coverage in your UEFN (Unreal Editor for Fortnite) project.
It always starts from a precise statement: preconditions, steps, player count and expected result. Then it inspects the affected files, configuration and the editor’s actual tool catalogue. A connected server does not mean a simulator or test runner exists, so it checks before picking a method.
How it works
It picks the smallest check that answers the question and keeps evidence levels apart:
- Static inspection of files and configuration;
- Compiler diagnostics, through BuildAll when the editor exposes it;
- A discovered automated test, run only through a tool with a known contract;
- Device properties and wiring reread in the editor;
- An observed client session, when one is accessible.
It covers the reported failure and its edge cases: duplicate events, zero and maximum values, join and leave, respawn, reset, several players. For each check it records the expected value, the actual value, the tool output and the file or asset involved. After a fix, it reruns the failed check and the smallest useful regression case.
When you allow it, it creates or adjusts scoped test files, with every write going through your approval. It pairs with the Verse Coder, which fixes, and the Reviewer, which reviews without changing anything. The orchestrator can hand it the verification part of a plan.
Example prompts
- Write the test checklist for a shop purchase.
- Coins sometimes get counted twice: reproduce the bug and find where it comes from.
- Check that the upgrade buttons are wired to the shop.
- Check the reward math at zero coins and at the maximum.
- After the Verse Coder’s fix, rerun the test that failed.
Tools and connections
- Connections
- UEFN, Blender (optional)
- Build
- BuildAll through UEFN’s MCP, when the editor exposes it
- Files
- scoped test files, when authorized
- Starting Skills
- pilot-testing, pilot-qa
- Verdicts
- PASS, FAIL or NOT RUN for each check
Limits
No simulation engine or automatic playtest is assumed to exist. A missing tool or unreachable gameplay gives NOT RUN, never PASS. The Tester never claims offline simulation, full coverage, performance measurements or a playtest from code inspection alone. It does not quietly refactor production logic, change project settings or spend generation credits to manufacture a test.
To see how it fits into the team, visit the AI agent team page, then pricing.
FAQ
Can it play my map for me?
No, there is no automatic playtest. It observes a client session only when one is accessible; otherwise the check is marked NOT RUN and it tells you exactly what to test yourself.
What do PASS, FAIL and NOT RUN mean?
PASS: the check ran and the result matched. FAIL: it ran and the result differed, with evidence. NOT RUN: it could not run, with the exact reason.
Can it fix the bug itself?
Its job is to prove. It can write scoped test files when authorized, but fixing the logic belongs to the Verse Coder; it then reruns the test to confirm.