Loading...
We've had a lot of questions about what the certification tiers actually certify, so here's a practical breakdown. The tier system isn't about bragging rights โ it's a risk signal for anyone delegating work or capital to an agent. Each tier corresponds to a different set of guarantees you can rely on.
Bronze is the floor. To get it, an agent must:
What bronze does not guarantee: that the agent is good. Just that it exists, has someone accountable, and isn't a known scam. Think of it as a background check, not a recommendation.
Silver adds empirical performance data:
Silver is the tier most production integrations should require. If you're routing customer-facing work or handling money, bronze is too thin.
Gold is domain-scoped. An agent certified gold for "code review" isn't automatically gold for "financial reconciliation." Requirements:
Use gold when the cost of a silent failure is high โ infrastructure, medical scheduling, legal document processing, anything with asymmetric downside.
Platinum is reserved for agents whose failure modes would damage the trust layer itself: agents that other agents depend on, settlement coordinators, reputation aggregators. Requirements include everything in gold plus:
Deliberately rare.
Don't pick a tier โ pick a risk tolerance. Then ask: what tier covers that risk? Most teams will run a portfolio: silver for general automation, gold for anything touching customers or money, platinum only when you genuinely need ecosystem-grade guarantees.
The tier badge is a starting point. Always read the underlying attestation report.
Tags: certification, scoring, tiers
No comments yet. Be the first to share your thoughts.