No. A test runs before release against fixtures, mocks, or a sandbox. Outcome verification runs after every production action against the real provider. A test proves the code can succeed. A verdict proves this action did.

AspectTestingOutcome verification
WhenBefore releaseAfter each production action
AgainstMocks, fixtures, a sandboxThe real provider, with a credential that you give the verifier
ScopeCode pathsOne operation and its outcome
ResultPass or failVERIFIED, PENDING, CONTRADICTED, FAILED, UNVERIFIABLE
ProofA test reportA signed receipt
Duplicates and stale readsOnly if you wrote the caseBlocked by negative postconditions and the stale-read window
the promptthe agenttool callStripe200 pendingthe agent reads the status code and stops“Done. The customer has beenrefunded $142.”an agent assertion is evidence level E0E0 is never sufficient on its ownNobody compares the claim with the record, so nobody finds the difference.
The gap that outcome verification closes, and no other tool reads.
In words
  • A person asks the agent to perform the action.
  • The agent calls the provider and reads a success response.
  • The agent reports the task as done.
  • The provider state is still not the promised outcome.

When is testing the right tool?

Use testing for catching regressions before release and pinning behaviour. Use outcome verification when an agent makes a side-effecting call and must not say "done" until the provider proves it.

Questions developers ask

Does Provely have tests of its own?

Yes. Every integration passes the conformance matrix before it ships. The cross-language fixtures prove that TypeScript, Python, and Rust agree. Read the benchmark.

Can I use verification in CI?

Yes. provely verify --wait exits with the verdict code, so a pipeline step passes only on VERIFIED.