Changelog
Follow up on the latest improvements and updates.
RSS
Simulations, a new capability of DigitalOcean Evaluations, lets you test a change through realistic multi-turn scenarios in an isolated sandbox on Harness Runtime, with the same configuration and Action Gateway tools your agent uses in production. Pull scenarios from Signals, generate them from your agent’s configuration, or describe one in plain language. An LLM judge scores every trajectory, and running baseline against candidate shows exactly what improved and what regressed.
Signals watches production traffic on Managed Agents and Inference Engine and automatically highlights potential problems across user sentiment, tool execution, and infrastructure, with nothing extra to instrument. Flagged sessions become evaluation scenarios in one click. Signals is opt-in, designed to redact PII before processing, and keeps traces within your DigitalOcean Team boundary.
Insights gives you one place to detect, investigate, and monitor what your agents, applications, and infrastructure are doing. Metrics, logs, and traces flow in automatically from supported products with no instrumentation to set up. Insights includes dashboards, explorers, a query builder with PromQL support, alerts, a public API, and role-based access control. Available at no additional charge during public preview.
Session lengths are now configurable to stay signed in based on your own needs.
- Stay signed in on your terms:Set as short as 1 hour on shared or high-risk devices, or as long as 30 days on devices you trust.
- No early cutoffs: Once set, your session lasts the complete duration from sign-in, whether you're active the whole time or step away and back. It won't end early just because you were idle.
- One setting, every device:Change it once and it applies across every browser and device you're signed into.
- Changes apply going forward: Updating your duration won't lengthen or shorten a session you're already in. It takes effect the next time you sign in.
Organization Owners and Admins can now see and manage resource limits for every team in one place, instead of checking each team one at a time.
- One view across every team:The new Resource limits tab in Organization settings shows usage and limits for every team's Droplets, Volumes, Databases, and Load Balancers in a single table.
- Instant limit adjustments:Owners and Admins can raise or lower any team's limits themselves, within that team's approved maximum, no ticket required. Need more than the approved maximum? Request the increase from the same screen.
- Availability:Live now for every organization, visible to Owners and Admins (hidden from Billers). Teams keep their existing per-team limits page, and once an org admin adjusts a team's limits, that team is directed to the admin to request any further change.
One plan covers the compute, inference, storage, and tools your agents use, and the discount applies automatically to all of it. Pro and Team plans give you 15%, and 20% discounts on your usage. When your plan credit runs out, you choose: keep going at list price or stop at the plan price. Run one agent or a hundred, no seat charges.
GPT-6.1 Sol is OpenAI’s upgrade to GPT-6 Sol, available now through Serverless Inference. It targets near-GPT-6 Astra performance on coding, computer use, and document-heavy work at a fraction of Astra’s per-token cost, running on the same DigitalOcean account, API, and invoice as the rest of the GPT-6 family.
Enhance your Kubernetes cluster security with network-level isolation. Isolated Worker Nodes allow you to run DOKS clusters with zero public IPv4 addresses assigned to your worker nodes, completely removing direct public exposure. Outbound traffic for node provisioning, control plane registration, and container image pulls is securely routed through a NAT Gateway in your VPC. You can start creating isolated clusters today via the DigitalOcean Cloud Console or API. Learn more →
Claude Sonnet 5.5 is Anthropic’s newest mid-tier model and a direct upgrade from Sonnet 5, with stronger everyday coding, sharper writing, and more polished documents. It’s available now via Serverless Inference at the same per-token pricing as Sonnet 5.
NVIDIA Nemotron™ 3 Diarization, an open-weight, streaming speaker diarization model, is now available through DigitalOcean Serverless Inference. It provides real-time speaker attribution for up to 8 speakers in a single stream—2x the speaker support of the previous generation—and works alongside existing ASR pipeline without replacing your transcription model.
Load More
→