DigitalOcean Launches Insights, Built-In Observability for Agents and Infrastructure

One backend for agents, apps and infrastructure

DigitalOcean says production agents span model calls, routing decisions, tools and the infrastructure underneath, so when a run slows or fails the evidence is scattered across separate logs, metrics, traces and consoles. Applications on DigitalOcean Kubernetes (DOKS) or App Platform face a similar problem, with teams often building their own integrations or adding a third-party observability tool before they can diagnose anything.

Insights is meant to remove that setup work. Telemetry from DOKS, App Platform, Managed Agents and other products flows in through built-in integrations. The service runs on OpenTelemetry, the open standard for observability data, so the data stays in an open, portable format. Because everything comes from the same platform, signals share dimensions such as resource, project and tag, which lets an anomaly in one signal lead to related ones. DigitalOcean says this is the same observability layer that already powers Managed Agents (public preview).

Tools and coverage

Beyond collecting telemetry, Insights includes several tools for working with it:

• Dashboards to assemble the signals that matter into a single view.
• Explorers to browse metrics, logs and traces, drill down to a single resource, and follow a trace from its root span to the call that failed.
• Query with a visual builder or direct PromQL, with time series, stat, gauge, bar or pie visualizations that can be saved to a dashboard.
• Alerts on any metric, with a configurable evaluation window and re-notification frequency, routed to existing notification channels.
• API for managing alerting and notification channels and querying metrics, logs and traces programmatically.
• Access control through role-based access control.

On the agent side, metrics and logs are available for Serverless Inference, Batch Inference, Inference Router, Managed Agents and Action Gateway. DigitalOcean says this lets users watch token throughput, latency, error rates and cost estimates along a request's path. Action Gateway also adds traces, so a single agent turn can be followed from the model call through each tool invoked. DOKS and App Platform report metrics into the same views, along with logs for DOKS.

Pricing, Monitoring users and what comes next

Insights is free during public preview, subject to the Insights Public Preview Terms and any usage or retention limits in the documentation. DigitalOcean expects to begin charging at general availability (GA) and says it will announce pricing before billing starts.

The company also plans to consolidate its whole observability experience under Insights around GA. Current DigitalOcean Monitoring users will see their existing metrics and alerts keep working until the transition. DigitalOcean says it will share details, including any required steps or pricing changes, before that happens.

During the preview, DigitalOcean plans to expand coverage, including custom telemetry from users' own applications and support for more products. The company notes that it expects but does not guarantee that public preview services perform for production-level workloads.

For teams already running agents or workloads on DigitalOcean, the pitch is a single place to trace a problem from a slow agent response to the tool call behind it, or from a spike on DOKS to the pod driving it, without switching tools. Users can open Insights in the control panel or read the docs to get started.