The Pulse
DigitalOcean Puts Agent Data Services Into Private Preview
DigitalOcean announced Agent Data Services and Agent Learning Services on October 12, 2026, positioning them as tools for collecting agent context and testing improvements. Agent Data Services is in private preview, while the learning produ
AI.info Team · Written with AI ·
DigitalOcean is pitching a path for agents to improve from their own operating history, but the most automated part of that pitch is not yet available for agent workflows. Announced October 12, Agent Data Services is in private preview, while the company’s Agent Learning Services are in public preview. DigitalOcean says it plans to extend its automated improvement loop to agent workflows later; for now, it says human approvals remain part of that process.
One catalog for documents, media and agent sessions
Agent Data Services is designed to process incoming documents, images, audio, video, structured records and agent activity, then make the resulting material available through a shared catalog. DigitalOcean’s data flows can transcribe, extract, redact, chunk and embed information as it arrives. The company says agents can then search that material by meaning, while developers can query the catalog through an API or SQL.
The service also collects behavioral data, including logs, events and agent traces. Its documentation says those events can be assembled into sessions, redacted and stored alongside multimodal content, so a team can query an interaction together with the documents or media involved. DigitalOcean presents that shared store as a way to reduce the custom ingestion and indexing work teams often build around separate sources.
Processing happens when data arrives, rather than every time an agent asks a question. DigitalOcean’s documentation says the service stores processing results for later retrieval, avoiding repeated model inference for the same material. The company says workspaces keep data, flows, catalogs and query endpoints isolated by team and region.
The “self-improving” claim has a boundary
Agent Learning Services comprises Insights, Signals, and Evaluations and Simulations. Insights records agent activity; Signals identifies sessions that may show failures or user frustration; and Evaluations and Simulations let teams test changes against conversations based on those sessions. The products connect operational evidence to testing, but that does not mean an agent rewrites or deploys its own behavior without review.
“We built Inference Engine so agents could think, Managed Agents so they could act, and now Agent Learning Services and Agent Data Services so they can improve, on one open cloud and on the customer’s own data.”
Vinay Kumar, Chief Product and Technology Officer, DigitalOcean
The distinction appears in the company’s own launch materials: DigitalOcean says its Inference Router already uses an automated loop to select models against cost, quality or a team’s evaluations. The company says it plans to bring a similar automated loop to agent workflows, augmenting current human approvals. That puts workflow automation on the roadmap, not among the capabilities the announcement says are available today.
Preview status limits production use
DigitalOcean makes Agent Data Services available only to accepted private-preview participants. The Learning Services products are in public preview, according to the company’s announcement. The data-service documentation describes access through the API and control panel during the private preview, and says there is no doctl, client-library or MCP-server support at this stage.
DigitalOcean’s preview terms carry a practical warning: Agent Data Services is provided as-is, without a service-level agreement or support commitment, and is not covered by the company’s standard compliance certifications. The company also says preview features, availability, pricing and timing may change or be discontinued. Those conditions make the launch an early test of a unified agent-data workflow, rather than a production commitment for teams with strict availability or compliance requirements.
The announcement’s central proposition is that agents can read company data and write back records of their work through one managed service. Whether that shared store helps teams improve agent behavior will depend on the quality of the data flows, the evaluation process and the human decisions around proposed changes. For now, Agent Data Services remains a private preview, and DigitalOcean says automated improvement for agent workflows is still ahead.