Documentation
OpenHands tasks in a sandbox
Coming soon. OpenHands tasks are in development and not yet tested in a live sandbox. They are not available yet.
Use case: delegate a bounded task
Once enabled, ask OpenHands to add tests for one module under a fixed per-task AI budget and workspace spending caps. Tasks need a supported workspace AI provider. Tasks work with OpenAI-compatible chat providers. Native Anthropic and Responses providers do not work. Provider keys stay outside the sandbox. When the budget runs out, the task stops. If a call's usage is unknown, Runzivo counts the full reserved amount. Review changes and test results before keeping them.
Task controls after enablement
Start the sandbox, open its Agent panel and select Start Agent Server. When task access is available, enter a prompt and select Submit task. Watch events, send a follow-up, pause, resume or cancel. Pause waits for the current model call. Cancel stops the server. Closing the panel leaves the task running. Restart the server for another task after completion.
Access and lifetime
Submit and steer need sandboxes:write and ai_chat:use. Only the task owner can steer, pause or resume. Sandbox writers can cancel. Task events need sandboxes:write. Readers see status only. Anyone able to run sandbox commands, including through sudo, can use the active task’s AI access up to its budget. Share command access only with trusted people and code. Sessions last at most one hour and end when the sandbox stops or expires.
Billing
No new AI charge or price is activated by this preview. Existing sandbox compute still applies. Installed Codex and Claude Code are separate tools with their own authorized logins.
The panel does not estimate remaining spend.