DigitalOcean has made Learning Services and Agent Data Services publicly available in preview. These offerings enable developers to develop, assess, and iteratively enhance AI agents within a single, AI-native cloud environment.
These new services augment DigitalOcean's existing Inference Engine and Managed Agents. They allow agents to gain knowledge from real-world interactions, pinpoint deficiencies, validate enhancements, and utilize proprietary company data, the company stated.
Learning Services integrates three functions: Insights, which documents agent activity; Signals, which pinpoints performance gaps and failures; and Simulations, which tests potential upgrades via realistic conversations before deployment.
The enhanced Inference Router, though distinct from Learning Services, is crucial to DigitalOcean's strategy for continuous agent refinement. It balances cost against quality, optimizes model selection per task requirement, and re-runs evaluations on newer model iterations.
Agent Data Services establishes a secure, managed data base for AI agents. It indexes and processes various data types—including logs, video, audio, agent sessions, and documents—making this information searchable via a unified catalog complete with access controls and governance.
Agent Learning Services operates atop Data Services. Signals writes sessions into Data Services, and the Optimizer retrieves its training and testing data exclusively where the customer gives explicit consent.
Jonathan Saring, co-founder of LakeOps, noted on LinkedIn, "The write-back loop is the part I find most interesting. Once traces, tool calls and outcomes feed evaluation or learning, they stop being 'just logs' and become production data."
Deepanshu Udhwani, a senior engineer, argued that the outcome logging aspect is undervalued compared to the data retrieval side discussed with agent data layers. He mentioned that many teams build effective retrieval mechanisms but neglect proper logging of results, meaning six months later, it is unclear which tool calls truly succeeded versus those that appeared adequate momentarily.
Keith Zhai, co-founder of TinyFish, summarized DigitalOcean's launch clip by stressing how Signals and Simulations can resolve a failed user session into a fix before numerous other users encounter the same problem. He added, "Signals and Simulations pull that session into a fix before the four-hundredth person hits the same wall. [...] Green tool calls next to abandoned users is still where most agent stacks break. The bottleneck is whether anyone can read the session the user actually lived."
DigitalOcean Insights, Signals, and Evaluations, including Simulations, are accessible in public preview. Inference Router optimization and Agent Data Services are currently in private preview.
This development reflects a broader industry shift toward operationalizing AI agents, moving beyond simple inference to incorporate feedback loops necessary for autonomous system maturation. Market participants are clearly focusing on the infrastructure required to govern and learn from production agent behavior.
