Remote MCP Support is now available
Remote MCP support is now live on DigitalOcean for 9 services: Accounts, App Platform, Databases, DigitalOcean Kubernetes, Droplets, Insights, Marketplace, Networking, and Spaces. This allows AI tools to connect without local binaries. Each service has a dedicated HTTPS endpoint (e.g., https://apps.mcp.digitalocean.com/mcp). Update your mcp.json configuration with the hosted endpoint and your API token for immediate, authenticated, zero-setup access with no local dependencies. All existing MCP tutorials remain valid.
Introducing the Gradient™ AI Agent Development Kit
The ADK, now in public preview, is a code-first SDK for building and operating production-ready AI agents across popular Python frameworks, including LangGraph, LangChain, CrewAI, and AutoGen. Built-in tracing, evaluations, and Knowledge Base support make it easier to move from local development to production. Enable the ADK public preview now →
New Anthropic models now available on the Gradient™ AI Platform
Claude Opus 4.1, Opus 4.5, and Sonnet 4.5 are now available on the DigitalOcean Gradient™ AI Platform for Serverless Inference and the Agent Development Kit, giving developers more powerful options for building and scaling AI applications. For details, see the Available Models page →
Evaluate your AI agents more effectively
We’ve redesigned agent evaluations in the Gradient™ AI Platform to make testing AI agents faster, clearer, and easier to adopt. The new experience introduces goal-oriented metric groups, example datasets, clearer error messages, and deep trace integration to help developers diagnose and improve performance quickly. Learn more in the announcement blog →
DigitalOcean Gradient™ AI Platform Knowledge Base Enhancements
The enhanced DigitalOcean Gradient™ AI Knowledge Base, now in public preview, lets developers create, manage, and query knowledge bases entirely in code. With direct API access, flexible ingestion, and citation-backed search, the enhancements makes it easier to power accurate, production-ready RAG workflows. Enable the KB enhancements public preview now →
Simplify Droplet Troubleshooting with the new SRE Agent in Private Preview
Introducing the Droplet SRE Agent, an AI assistant designed to help with smarter Droplet management. Spot CPU spikes, investigate issues, and receive AI-generated recommendations—all in the Cloud Console. Sign up for the private preview to try it early and share feedback to help shape its development.
Usage-based backups to meet aggressive RPOs and slash costs
Our new usage-based backup service is now generally available, allowing you to schedule high-frequency backups for your Droplets and GPUs every 4, 6, or 12 hours to meet strict Recovery Point Objectives (RPOs). This feature comes with transparent, consumption-based billing, meaning you only pay for the amount of restorable data you use, drastically reducing costs for high-change environments like CI/CD pipelines and AI/ML data protection.
Spaces cold storage for infrequently accessed data and archives
Spaces cold storage is now generally available as a low-cost, S3-compatible solution for petabyte-scale datasets that are rarely accessed. Store your backups, archives, and application logs for just $0.007/GiB per month. Our simple pricing includes instant retrieval and one retrieval per month at no extra cost, providing predictable, cost-effective scaling for all your retention and compliance needs.
Network File Storage (NFS) for high-performance AI workloads
We launched Network File Storage, a managed, high-performance shared file system optimized for AI/ML. It enables concurrent access to shared datasets for multi-node GPU/CPU workloads. Provisioning is simple, starting at cost-effective 50 GiB increments, and it meets high-throughput, low-latency demands. This NFS solution is now available in our ATL1 and NYC2 data centers.
Per-second billing for Droplets to slash costs for ephemeral workloads
Effective January 1, 2026, we're moving to per-second billing (with a 60-second minimum) for Droplets. This offers truly granular cost control, drastically reducing spending on short-lived workloads, such as running batch jobs and automated testing. The monthly cap remains in place for budget predictability on sustained tasks.
GPU Droplets accelerated by NVIDIA HGXTM B300 are coming soon
The NVIDIA Blackwell Ultra accelerated computing platform represents a leap forward in AI reasoning. Designed for both training and inference, NVIDIA HGX B300 offers substantial improvements in computational power, memory bandwidth, and energy efficiency compared to previous generations.
Introducing the custom date range billing view, a new cost insights tool
All DigitalOcean customers now have access to the new custom date range billing view. This new feature is accessible in the Billing Console by going to Billing → Insights. Additionally, you can download your report as a CSV so you can view your granular billing insights offline. This new tool gives you clearer, more granular visibility into your cloud costs (segmented into daily, weekly, and monthly spend by product) making budgeting and forecasting far easier.
Session Management: Cloud Console sessions will now time out after 12 hours. We’ll give you a heads-up at 10 minutes and 1 minute before it ends. Just sign back in on any tab to pick up right where you left off.
DigitalOcean Kubernetes improvements: DigitalOcean Kubernetes (DOKS) now officially supports Kubernetes v1.34, which supports specialized workloads, particularly AI/ML workloads, offers operational resilience and cost optimization with in-place pod resizing. This version also improves both the security and the developer experience by leveraging updated API functionalities/integrating enhanced container isolation. This update will be automatically applied to your Kubernetes environment.
Improved migration tooling in the cloud console: In December, we improved migration tooling in the Cloud Console, making it easier to move data into Managed Databases with clearer guidance, validation, and a streamlined workflow. We also enhanced reliability and monitoring across our database offerings, including automatic restarts for MongoDB single-node clusters, better alerting for resource and performance issues, and improved single-node uptime. Additionally, MySQL incremental backups were introduced, enabling faster, more efficient backups and giving customers greater control over their data protection.
