DigitalOcean Agent Learning and Data Services
DigitalOcean announced Agent Learning and Data Services on 12 October 2026: the company says the services turn agent interactions into better business outcomes by recording every step with Insights, flagging the sessions that matter with Signals, and testing changes before real users with Evaluations and Simulations
Tool Overview
Details, capabilities and usage guidance
Features
- ✓ Insights: DigitalOcean says it records every step an agent takes and brings infrastructure, application and agent observability into one place, with telemetry from DOKS, App Platform and Managed Agents flowing in through built-in integrations and no additional instrumentation to set up
- ✓ Signals: DigitalOcean says it flags the sessions that matter - where the agent stalled, looped, lost the user or failed
- ✓ Evaluations and Simulations: DigitalOcean says these build realistic conversations from flagged sessions so changes can be tested ahead of real users
- ✓ Recursive self-improvement: DigitalOcean says the loop is already applied to models - the enhanced Inference Router picks the best-scoring model for each task optimised for cost, quality or the team's own evaluations, and re-runs those evaluations as new models ship
- ✓ A companion data layer: DigitalOcean says Agent Data Services is a fully managed data layer built for agents. Insights, Signals, and Evaluations and Simulations are available today in public preview, while the Inference Router optimizer and Agent Data Services are in private preview with access on request
How to Use
- Read the official DigitalOcean blog posts for Insights, Signals and Simulations and Agent Data Services to confirm each capability's preview status
- Use Insights, available in public preview, to connect telemetry from DOKS, App Platform or Managed Agents and see how agents really behave
- Use Signals to find the sessions where agents stalled, looped, lost the user or failed, turning that evidence into a proposed change
- Test the change ahead of real users with Evaluations and Simulations built from flagged sessions, and request private preview access for the Inference Router optimizer or Agent Data Services as described officially
FAQ
What are DigitalOcean Agent Learning Services?
DigitalOcean says Agent Learning Services turn agent interactions into better business outcomes and, together with Agent Data Services, a fully managed data layer built for agents, form part of its modern agentic stack - giving teams what the company calls one platform to run, learn and improve agents.
What problem do they solve?
DigitalOcean says Insights records every step an agent takes and Signals flags the sessions that matter, such as where an agent stalled, looped, lost the user or failed, turning that evidence into a proposed change that Evaluations and Simulations then test ahead of real users.
Which parts can I use now?
DigitalOcean says Agent Learning Services, meaning Insights, Signals, and Evaluations and Simulations, are available today in public preview, while the Inference Router optimizer and Agent Data Services are in private preview, with teams able to request access as described officially.
What does the Inference Router do?
DigitalOcean says the newly enhanced Inference Router picks the best-scoring model for each task optimised for cost, quality or the team's own evaluations, and re-runs those evaluations as new models ship - applying recursive self-improvement to models first.
How is this different from generic observability tooling?
DigitalOcean says telemetry from DOKS, App Platform, Managed Agents and more flows in through built-in integrations with no additional instrumentation to set up, so teams see infrastructure, applications and agents in one place instead of assembling cross-system integrations themselves.
How can I try it?
DigitalOcean says it will demonstrate the services live at its Open Intelligence Summit in San Francisco on 13 October and that teams can request private preview access through its pages. A practical path is to start with Insights and Signals in public preview before evaluating deeper automation.