--- title: Managed Agents (public) description: DigitalOcean's cloud platform for running and scaling AI agents, and governing the actions they take. product: Managed Agents url: https://docs.digitalocean.com/products/managed-agents/ last_updated: "2026-10-06" --- > **For AI agents:** The documentation index is at [https://docs.digitalocean.com/llms.txt](https://docs.digitalocean.com/llms.txt). Markdown versions of pages use the same URL with `index.html.md` in place of the HTML page (for example, append `index.html.md` to the directory path instead of opening the HTML document). # Managed Agents (public) Managed Agents, available in [public preview](https://docs.digitalocean.com/platform/product-lifecycle/index.html.md#public-preview) for all users, bring together hardware-isolated microVM sandboxes and governed tool access so you can scale agentic work in the cloud without building and maintaining any underlying infrastructure. Execute code from an existing application, run complete agents with durable sessions, or connect agents to the tools they need to act. Managed Agents includes two services that you can use independently or together, with each scaling to meet your specific needs: [DigitalOcean Harness Runtime](https://docs.digitalocean.com/products/managed-agents/agent-harness-runtime/index.html.md): DigitalOcean Harness Runtime provides managed, hardware-isolated microVM sandboxes with built-in tools such as Chromium to run agents and execute arbitrary code. Rich lifecycle APIs preserve conversational history and working state across sessions, letting you pause, resume, and fork work to control costs and adapt to the nonlinear nature of agentic workflows. Scale complete agents such as Claude Code or use sandboxes independently for code execution, all through the same service. - **Execute code from any workflow.** Keep your agent or application wherever it runs today and use Harness Runtime to execute code in isolated sandboxes. Run generated Python, test changes, process files, or distribute independent tasks across sandboxes in parallel. - **Run the agent itself.** Launch harnesses such as Claude Code, Codex, and OpenCode inside microVM sandboxes with the filesystem, dependencies, and local tools they need to work. Harness Runtime manages the session lifecycle so you can run agents in parallel, preserve their work, and pause, resume, or branch sessions as tasks evolve. - **Control how work continues.** APIs that let you pause sessions, resume with context intact, or fork from a checkpoint to explore different approaches in parallel. These controls, together with active CPU billing, help you manage costs and adapt execution to the way agents actually work. Both scenarios use the same service and lifecycle controls, so you can scale code execution and complete agents without maintaining separate infrastructure for each. See [What You Can Build with Managed Agents](https://docs.digitalocean.com/products/managed-agents/agent-harness-runtime/details/what-you-can-build/index.html.md) for example use cases. [Action Gateway](https://docs.digitalocean.com/products/managed-agents/action-gateway/index.html.md): Action Gateway gives agents access to 16,000+ tools through a single managed MCP endpoint. The catalog includes DigitalOcean-maintained tools for web search, web fetch, browser automation, and DigitalOcean infrastructure management, alongside connectors for platforms such as GitHub, HubSpot, Stripe, and Snowflake. Teams can also extend the catalog with their own MCP servers and internal tools. Together, these services provide a managed foundation for running agents, preserving their work, and governing their access to tools. You can focus on making agents more useful while DigitalOcean manages the execution, persistence, and tool infrastructure underneath. ![Managed Agents architecture. A Managed Agents boundary contains Harness Runtime, which runs Claude Code, Codex, OpenCode, Hermes, LangGraph, CrewAI, or your own OCI image unmodified in a microVM, and Action Gateway, which fronts 16,000 or more tools behind one governed endpoint. Tool calls pass between them, and both emit events to Signals. Outside the boundary, Managed Agents connects to Inference Engine for model calls, Storage for reads and writes, and Data services for queries.](https://docs.digitalocean.com/screenshots/managed-agents/architecture.7fbd90d93e2954b7f5ef84af0622788f76f578ed0596e7db5f7fef31f973ee2e.png) For a walk-through of each component and the DigitalOcean services around them, see [Managed Agents Architecture](https://docs.digitalocean.com/products/managed-agents/agent-harness-runtime/concepts/architecture/index.html.md). [DigitalOcean Harness Runtime](https://docs.digitalocean.com/products/managed-agents/agent-harness-runtime/index.html.md): Run AI coding agents in isolated DigitalOcean sandboxes with session lifecycle, permissions, and a YAML environment spec. [Action Gateway](https://docs.digitalocean.com/products/managed-agents/action-gateway/index.html.md): Give agents access to tools, APIs, and connectors through a managed MCP endpoint or SDK without putting credentials in the model context. ## Latest Updates ### 6 October 2026 - [DigitalOcean Harness Runtime](https://docs.digitalocean.com/products/managed-agents/agent-harness-runtime/index.html.md) now supports CrewAI agents. Package a CrewAI repository, declare kickoff inputs, connect a model and Action Gateway tools, and deploy your crew in an isolated sandbox. For more information, see [How to Run a CrewAI Agent on DigitalOcean Harness Runtime](https://docs.digitalocean.com/products/managed-agents/agent-harness-runtime/how-to/run-crewai-agent/index.html.md). - Harness Runtime Bring-Your-Own-Template (BYOT) registration now requires a recommended base template (`codex-base`, `opencode-base`, `claude-code-base`, `hermes-base`, or `langgraph-base`). Recommended bases provide stronger isolation for custom templates than previously supported bases (`coding-codex`, `coding-opencode`, `coding-claude-code`, `coding-hermes`, and `langgraph`). The legacy bases remain available for existing custom templates and for rebuilds of those templates. Support for them in BYOT will be removed in a future release. Sessions that use the legacy base templates as adapter default templates continue to work as usual. For more information, see [How to Use a Custom Sandbox Template (BYOT)](https://docs.digitalocean.com/products/managed-agents/agent-harness-runtime/how-to/use-custom-template/index.html.md). ### 1 October 2026 - You can now prepay for [Serverless Inference](https://docs.digitalocean.com/products/inference/how-to/si-overview/index.html.md) and [Managed Agents](https://docs.digitalocean.com/products/managed-agents/index.html.md) with the Inference and Agents balance, which no other products draw from. Add $5 to $500 at a time or turn on auto-reload from the **Billing** page. Funds do not expire. You can also pay for eligible usage at a discount with a monthly plan, now in [public preview](https://docs.digitalocean.com/platform/product-lifecycle/index.html.md#public-preview). See [Inference and Agents Balance](https://docs.digitalocean.com/platform/billing/pay-bills/index.html.md#inference-agents-balance) and [How to Pay for Harness Runtime](https://docs.digitalocean.com/products/managed-agents/agent-harness-runtime/how-to/pay-for-harness-runtime/index.html.md). ### 21 September 2026 - [Managed Agents](https://docs.digitalocean.com/products/managed-agents/index.html.md) is now available in [public preview](https://docs.digitalocean.com/platform/product-lifecycle/index.html.md#public-preview) for all users. Managed Agents is a set of services that gives every agent its own microVM execution environment and governed tool access so that developers can scale agentic work without having to build and maintain infrastructure themselves. Developers can pick from agent harnesses that they already use (like Claude Code or agents built with frameworks like LangGraph and Hermes), securely connect it to more than 16,000 tools and have their agents do meaningful work in seconds, from the Control Panel or the DigitalOcean CLI. Teams can also register custom OCI images as sandbox templates with Bring-Your-Own-Template (BYOT). [DigitalOcean Harness Runtime](https://docs.digitalocean.com/products/managed-agents/agent-harness-runtime/index.html.md) combines the functionality of a lightweight microVM, built-in tools like chromium and a coding sandbox needed by agents to do work. The product also offers rich lifecycle APIs that persist conversational history and working state across sessions, along with pause/resume/fork semantics so that developers can control costs and adapt workflows to the nonlinear quirks of agentic work. [Action Gateway](https://docs.digitalocean.com/products/managed-agents/action-gateway/index.html.md) gives agents governed access to 16,000+ tools through a single managed MCP endpoint. This includes DigitalOcean-maintained tools such as Web Search, Web Fetch, Browser Automation, and DigitalOcean infrastructure management APIs, along with connectors for platforms like GitHub, HubSpot, Stripe, Snowflake, and more. Teams can also extend the catalog with their own MCP servers and internal tools. For more information, see [the full release notes](https://docs.digitalocean.com/release-notes/index.html.md).