FAQ

Is Cloud Agents just a chat wrapper?

No. Cloud Agents hosts real agent runtimes with workspace files, terminal access, resumable sessions, provider settings, skills, and optional channel connections.

Which agent should I choose first?

For repository work, start with Claude Code or Codex. Choose Gemini CLI if your workflow already depends on Gemini. Use Hermes Agent or OpenClaw when you need a framework-style agent for connectors, services, scheduled jobs, or product workflows.

Do I need my own model keys?

You can bring your own provider keys. Some workspaces may also have managed model access. The creation flow shows the options available to your account.

What is the difference between a stateful sandbox and a persistent container?

A stateful sandbox is best for interactive coding and task work. It can pause and resume while keeping the workspace state.

A persistent container is best for always-on work such as services, connectors, or scheduled workflows.

Can I run agents on my own machine?

Yes. Use Settings -> Local daemons to issue a token, then register the machine with the nca CLI. Local daemons are useful when the agent needs access to local files, hardware, or a private network that is not available from the cloud workspace.

Can I connect one agent to multiple channels?

Yes. Create one channel per external bot or app. Keep labels clear so you know which team, room, or workflow each channel serves.

Where should I put secrets?

Put model keys in Settings -> Model providers and channel tokens in Settings -> Channels. Do not paste secrets into chat prompts, files, issue descriptions, or public logs.

Can agents make mistakes?

Yes. AI-generated work can be wrong or incomplete. Review important outputs, inspect code changes, and run tests before relying on an agent’s work.

How do I reduce unexpected costs?

Use clear prompts, stop runs that are going in the wrong direction, review Settings -> Usage, and confirm the agent is using the intended provider and model.

How do I get support?

Contact support from your Netmind workspace or email customer-support@netmind.ai. Include the agent name, approximate time, and a short description of what happened. Do not include secrets.