DigitalOcean

DigitalOcean has launched 'Managed Agents' into public preview, utilizing a Firecracker microVM-based architecture. This system enables the checkpointing and restoration of running processes, memory states, and filesystem contexts for agentic workloads.

DigitalOcean has introduced the Inference Router in Public Preview, which automatically routes API requests to the optimal model based on task-specific benchmarks. This feature eliminates the need for manual routing rules and redeployment when swapping between frontier and open-source models.

DigitalOcean has launched Action Gateway to govern external API calls made by AI agents within Managed Agents environments. The system decouples provider credentials from model context and enforces centralized policy checks for tool execution.

DigitalOcean has updated its Inference Router to automatically route API requests to the most performant model based on internal benchmarks for software engineering, agentic workflows, and general reasoning. The system now dynamically selects models from the DigitalOcean Model Catalog, including Anthropic, OpenAI, Kimi, GLM, and DeepSeek, without requiring manual routing rules or configuration.

DigitalOcean has launched Agent Data Services in private preview to provide a managed platform for ingesting, indexing, and querying multimodal data for AI agents. The service integrates automated data flows, a unified data catalog, and SQL/API-based retrieval to replace fragmented data pipeline infrastructure.

DigitalOcean launched 'Signals' for automated production traffic triage and 'Simulations' for multi-turn scenario testing within Agent Learning Services. These tools integrate directly with DigitalOcean Managed Agents, Inference Engine, Harness Runtime, and Action Gateway.

DigitalOcean has launched DigitalOcean Insights, a native observability platform that aggregates metrics, logs, and traces from DOKS, App Platform, and Managed Agents. The service utilizes OpenTelemetry standards to provide correlated telemetry without requiring manual instrumentation or third-party tool deployment.

DigitalOcean has launched MicroVMs, a new compute primitive based on Firecracker technology that supports automatic pausing, stateful resuming, and hardware-level isolation. This service is designed for AI agent workloads that require ephemeral, bursty compute environments with strict security boundaries.

DigitalOcean has launched MicroVMs in public preview, utilizing the Firecracker virtualization technology to provide isolated, hardware-virtualized compute environments. The service introduces a pause-and-resume capability that allows users to stop compute billing during idle periods while maintaining the VM's memory and process state.

DigitalOcean introduced a centralized Resource Limits tab for organization-wide management of Droplets, Volumes, Databases, and Load Balancers. Additionally, users can now configure individual session durations ranging from 1 hour to 30 days.

DigitalOcean has launched 'Agent Droplets,' a bundled service that integrates managed runtime, serverless inference, persistent storage, and access to over 16,000 tools. This offering replaces fragmented, multi-vendor billing with a single, predictable monthly subscription model for AI agent infrastructure.

DigitalOcean has launched MicroVMs, a new compute service based on Firecracker technology that provides hardware-level isolation for container images. The service introduces automatic pausing and stateful checkpointing, allowing workloads to resume from memory and file states after idle periods.

DigitalOcean released substantial architectural upgrades, introducing enhanced API integrations and updated workflow tooling for its Developer Tools ecosystem.

DigitalOcean has introduced 'DigitalOcean Managed Agents' to decouple sensitive credentials from agent execution environments. The platform now intercepts service requests to authenticate calls externally and enforces network-level egress controls based on NVIDIA OpenShell’s open policy schema.

DigitalOcean has launched 'Managed Agents' into public preview, providing a unified runtime environment for AI agents. The platform integrates sandbox execution, model inference, and data storage into a single service architecture.

Following a successful private preview, we’re thrilled to open DigitalOcean Managed Agents to everyone. Teams can now deploy their preferred agent harness (like OpenCode, Codex CLI) or bring their own, connect agents to ...

DigitalOcean has launched the General Availability of Managed Databases Advanced Edition for MySQL and PostgreSQL. The release introduces integrated proxy architecture for sub-3-second failover and a 46% reduction in storage costs to $0.115 per GiB/month.

Setting the stakes In early July, security researcher Hyunwoo Kim discovered Januscape (CVE-2026-53359), a flaw in KVM’s handling of nested ...

Intelligence was yours. A founder’s intellectual property, business logic, differentiation, their intelligence, was once theirs alone. You b...

Omarchy has migrated its AI agent pipeline infrastructure to DigitalOcean's cloud platform. This transition leverages DigitalOcean's compute resources to support the execution of Omarchy's automated agent workflows.

DigitalOcean introduced its fully managed GenAI Platform, enabling developers to build, fine-tune, and deploy AI agents and LLM applications using open-source models like Llama 3 and Mistral. The platform integrates serverless endpoints, retrieval-augmented generation (RAG) capabilities, and built-in vector search without requiring manual GPU cluster setup.

DigitalOcean has introduced its GenAI Platform, enabling developers to deploy serverless inference endpoints and fine-tune open-source models like Llama 3 and Mistral. Powered by NVIDIA H100 Tensor Core GPUs starting at $2.69 per hour per GPU, the platform provides scalable infrastructure for generative AI applications without requiring manual cluster setup.

DigitalOcean has launched v5 Droplets, a new generation of virtual machine instances. These instances are specifically architected to provide granular sizing options tailored to specific workload requirements.

Agent sessions that outlive your laptop...

Setting the stakes...

A warm cache can be more valuable than a cheaper model...

Hardware selection and implementation...

DigitalOcean has launched 'model synthesis', a server-side tool integrated into the DigitalOcean Inference Engine. This tool provides performance benchmarks exceeding Fable 5 while reducing operational costs by 50%.