DigitalOcean previews Learning Services and Agent Data Services

On November 6, 2026, DigitalOcean announced the public preview of Learning Services and Agent Data Services, a pair of offerings that let developers evaluate and continuously improve AI agents on top of their own data within DigitalOcean's AI-native cloud platform. The services build on the Inference Engine, which the company describes as what lets agents think, and Managed Agents, which let them take actions. With the new offerings, DigitalOcean positions a full loop in which agents learn from real-world interactions, surface weaknesses, test candidate improvements, and use enterprise data throughout, all running on a single open cloud platform rather than stitched together from separate tools.

Learning Services groups three capabilities under one roof. Insights records agent activity so teams can see what agents actually did. Signals automates the identification of failures and performance gaps, turning telemetry into concrete improvement targets. Simulations lets developers run realistic conversation simulations before deployment, testing potential changes without putting them in front of real users. A companion update to the Inference Router brings dynamic model selection based on task requirements, and lets teams reuse the same evaluation cases when they evaluate a new model version, which simplifies cross-model consistency checks.

The companies describe the target audience as AI engineers, MLOps engineers, and cloud-native platform architects. As of November 2026 the services are in public preview, with no general availability date announced yet.