Convoy

Your agent just embarrassed you in front of all your users? You need to ship in phases and use Convoy to auto-rollback on problems. This is how I did it in 60 seconds.

Introducing Convoy - Phased rollouts for AI agents. - Deploy a prompt change - Ramp to 5% traffic - Auto-rollback if quality goes down

Evals need to see day light. They deserve to see the real world.

Introducing: A/B testing for AI agents Test changes to: - models - prompts - tools Against real user traffic. Convoy routes a small slice of requests to the new version, evaluates the results, and automatically promotes or rolls back the change. @ConvoyLabs

Announcing Convoy @ConvoyLabs No matter how many evals you write, you still end up shipping broken agents to production. That’s the problem with evals. They don’t see real users. Convoy makes experimentation a first-class part of the agent delivery cycle. Ship to 5%. Learn from real user behavior. Auto-rollback when quality regresses. Convoy is live.