List the requirements you actually asked for
Break the original request into up to four discrete requirements, in your own words.
AGENT OUTPUT FULFILLMENT CHECK
Agent QA
List the original request's key requirements, mark how well each was met by the agent's output, and get a calculated request-fulfillment score.
- Uses your inputs in a transparent calculation
- Instant result with practical next steps
- No signup required
Planning guidance only. Validate important decisions with customer evidence and your delivery team.
YOUR INPUTS
Build your result
Complete every field. The result updates only when you choose Calculate.
How it works
Mark each one against the agent's output
For each requirement, mark it fully met, partially met, or not met based on what you observed after reviewing the change.
Get a calculated fulfillment score
Fully met requirements count in full, partially met count at half weight, giving a transparent 0-100 score plus which requirements to send back.
Continue learning: AI-assisted MVP engineering: what should humans decide · When an AI-generated demo needs engineering review
Frequently asked questions
Does this automatically test or validate my agent's output?
No. This tool computes a fulfillment score entirely from the checklist statuses you mark after reviewing the output yourself. It does not run tests, execute code, parse commits or pull requests, or connect to any coding agent — you supply the judgment, it does the scoring arithmetic.
Why only up to four requirements?
Four keeps the check fast and forces you to name the requirements that actually matter, rather than an exhaustive spec. For a larger task, run the check per logical group of requirements instead of cramming everything into one pass.
What does "partially met" mean in practice?
The agent addressed the requirement but not completely — for example pagination works but the page number doesn't persist on refresh. It counts at half weight so it moves the score without being treated the same as fully done.
Is a high score a guarantee the code is correct?
No — the score reflects your own assessment against the requirements you listed, not an automated correctness check. Pair it with an actual code review, especially for anything touching sensitive data or production.
How is this different from a QA test suite?
A test suite runs automated assertions against real code. Agent QA is a structured self-check you fill in by hand after reviewing an agent's output, meant to catch "it looks done but didn't actually meet the ask" before you close the task.
How We Compare
| Feature | MVPHub | Replit | Lovable |
|---|---|---|---|
| Requirement-by-requirement fulfillment score | Included | Not included | Not included |
| Flags specific unmet requirements to send back | Included | Not included | Not included |
| Automated test execution against output | Not included | Limited | Limited |
| Autonomous agent code generation | Not included | Included | Included |
Replit and Lovable can generate and sometimes self-test agent output, but neither prompts you through a structured requirement-by-requirement fulfillment check against your own original ask. Agent QA gives a fast, honest self-assessment before you close the task as done.
Related tools
Embed this tool
Add this tool to your site with the canonical iframe below. It remains hosted and maintained by MVPHub.