Managed Agents runs now show step details, outputs, and new configuration options

The Managed Runs page now shows run details, including step durations and logs. Filter runs by provider and status, and select a run to investigate failures. Organization owners can also stop active runs.

Download saved files from the Outputs section after a run finishes. To make files available, have the agent save them to /workspace/outputs/.

Each Managed Agents connection now has its own run settings, in one place under Settings > Managed Runs. They cover:

• Run secrets, including secrets that Depot sends only to hosts you list. Those secrets never enter the sandbox.
• A setup script.
• GitHub access through Depot's Code Access app.
• MCP servers that runs can use.

In runs from Claude Code and OpenAI connections, the depot CLI is now signed in as your organization, so the agent doesn't need its own Depot token.

See Managed Agents for setup and limits.