As reported on November 16, 2026, DigitalOcean had launched a public preview of DigitalOcean Managed Agents, a cloud service for running AI agents. It combines Harness Runtime, which manages agent execution and state, with Action Gateway, which controls access to external tools. DigitalOcean also describes the offering as including serverless AI inference.
DigitalOcean says developers running agents on conventional virtual machines must otherwise manage tasks such as preserving context and artifacts, coordinating parallel work, securing tool access and keeping capacity available for quick starts.
Harness Runtime runs supported agents in isolated microVMs. DigitalOcean says it preserves conversation history and work state across sessions, which can be paused, resumed or forked. It automatically pauses an idle session when no LLM or tool call is in progress. Supported options include Claude Code, Codex CLI, OpenCode, Hermes and custom agents built with LangGraph. Developers can run sessions in parallel and connect them to internal services without exposing those services to the public internet, according to the company.
Action Gateway provides a unified Model Context Protocol (MCP) endpoint for external services. The report says more than 16,000 tools are available, including web search and fetching, browser automation, DigitalOcean infrastructure APIs, and connectors for GitHub, HubSpot and Stripe. DigitalOcean says the gateway provides per-agent permissions, required human approval for sensitive actions, and rate limits, retries, backoff and timeouts.
Commenting on the launch, Dair.ai builder and former Meta AI engineer Elvis Saravia welcomed its focus on performance and cost. Seaotter founder and platform architect Ryan Martin highlighted idle-session pausing and governed tool access as an approach to operating agents at scale. These are their assessments, not reported performance or cost measurements.
The report also compares the launch with Docker Cloud Sandboxes, noting overlap in microVM isolation and persistence. It characterizes Docker’s offering as more focused on developer workflows between local and cloud environments, and DigitalOcean’s as more focused on production infrastructure and tool orchestration.