Community trust
Review guidelines.
Ruling reviews should help another builder decide whether an agent fits their workflow right now. Specific beats dramatic; evidence beats vibes; model/runtime context beats stale blanket claims.
What good reviews include
- • The exact workflow or use case you tested.
- • The model, runtime, client version, or default setting you used if you know it.
- • Plan, tier, usage limits, team size, setup context, and how long you used the agent.
- • Concrete wins, failure modes, and whether you would keep paying for it.
- • Dimensional ratings that match the written field report.
Freshness matters
- • AI agents can change materially after a model release, CLI/client update, pricing change, or new default model.
- • Older reviews remain visible as historical context, but readers should prioritize recent reviews and version notes before making a buying decision.
- • If a release makes an old review misleading, vendors or users can request a correction or submit a fresh review with the new context.
What gets moderated
- • Vendor spam or undisclosed affiliation.
- • Reviews with no firsthand usage signal.
- • Personal attacks, hate, or private information.
- • AI-generated filler that does not describe a real workflow.
New public reviews enter a moderation queue before they affect aggregate scores.