DigitalOcean
The AI-Native Cloud — from silicon to agents in one stack.
5.0•86 reviews•1.2K followers
The AI-Native Cloud — from silicon to agents in one stack.
5.0•86 reviews•1.2K followers
DigitalOcean's AI-Native Cloud — five layers, one integrated stack, open at every layer, zero lock-in. Inference Engine with 65+ models. Inference Router auto-routes 90%+ of calls to OSS for 60%+ savings. Managed Agents, Knowledge Bases, and serverless / dedicated / batch inference. 25-40% lower than hyperscalers. Trusted by Character AI (800B tokens/day), Hippocratic AI (8M calls @ 400ms), and many others. From a single line of code to scaling AI in production — start here, scale here.
This is the 20th launch from DigitalOcean. View more
DigitalOcean Managed Agents
Launching today
DigitalOcean Managed Agents runs AI agents in the cloud. Harness Runtime gives each session its own Firecracker microVM that can pause, resume and fork, and runs Claude Code, Codex CLI, OpenCode, Hermes, LangGraph, CrewAI or any container. Action Gateway gives agents 16,000+ tools through one MCP endpoint, with credentials kept from the model and approvals for risky actions. Public Preview, from $5 trial.
Free Options
Launch tags:Artificial Intelligence
Launch Team
Viktor.comAn AI coworker that actually does the work
Promoted
DigitalOcean Managed Agents is a cloud platform for running AI agents in production, with isolated sessions and governed tool access.
Problem: Agents that run on a laptop stop when it sleeps, and moving them to cloud VMs means building session persistence, credential handling and tool integrations yourself. Every tool also needs its own credentials, permissions and retry logic, with no clear way to trace what the agent did or what it cost.
Solution: Managed Agents has two parts, both in Public Preview. Harness Runtime runs each agent session in its own Firecracker microVM that can pause, resume and fork. Action Gateway gives agents governed access to tools through one managed MCP endpoint.
What makes it different: It isn’t tied to one agent or model provider. It runs Claude Code, Codex CLI, OpenCode, Hermes, LangGraph, CrewAI, or any agent packaged as a container image. Action Gateway keeps credentials away from the model and sandbox, and can require human approval for sensitive actions. One plan and one bill cover the runtime, sandbox, model and search API.
Key features:
Dedicated microVM per session, with parallel sessions and subagent workflows
Pause and resume without losing context, and CPU and memory charges stop while paused
Checkpoint a session and fork new ones from it
16,000+ tools through one MCP endpoint, with 500+ third-party providers
Per-tool permissions, per-user OAuth and human approval for sensitive actions
Scheduled and webhook triggers
Structured event logs with token usage and cost visibility
Who it’s for: Developers and teams running coding agents or automation in the cloud, and companies building agent-powered products who need isolation and safe tool access underneath.
New customers get a $5 trial credit, and paid plans start at $50 a month.