Cloud Agents
Roo Code Cloud Agents can run entire tasks on their own in the cloud, often automatically, so you're not limited to typing instructions in your IDE.
What Roo Code Cloud Agents are
Hosted, headless Roo-backed agents who:
- Run tasks in cloud containers independent of your local IDE (you don't even need to use the Roo Code VS Code extension in order to use Cloud agents)
- Have focused, optimized prompts for specific needs
- Stream task history in the web app
- Fit into your broader workflow, outside of the editor
Agents Types
Cloud Agents come in different types optimized for specific workflows. Each uses your chosen model and API key, running in the cloud so work continues without your IDE.
PR Rooviewer
Automatically reviews GitHub pull requests using your model, posts feedback, and tracks costs—triggered by webhooks or run on demand.
Coming soon
We have more agents in the works, such as:
- PR Fixer
- String Translator
- Security Auditor
- Generalist
How billing works
- Cloud Agents require an active paid plan.
- Cloud runtime is billed in credits per agent‑hour (we charge you while the agents are working, that's it)
- Model inference is billed by your provider (bring your own API keys). We don't mark inference costs up and are provider-agnostic
- You can track usage and credits in Analytics; switch the metric to credits.
Getting started
- Sign up for a Roo Code Cloud account
- Upgrade to a paid plan.
- You'll get free credits to get started
- Create your first agent.
FAQ
-
Can I run multiple agents?
- Yes. Each agent incurs its own runtime credit usage while running.
-
Do I need to use a specific model?
- No. We recommend a few models we've tested and found to work well, but you're free to use absolutely any model your provider offers.
-
Where do I see costs?
- Track runtime credits and usage in Analytics; see Billing & Subscriptions for plans and credits.