remoteagent.online  the open-source control plane for secure, cost-controlled AI agents

Enable secure mode for a locked-down agent

Turn on secure mode to run an agent with read-only shell, workspace-only files, and no network. The strictest profile available.

Updated 2026-08-22Reading time 3 min

What secure mode changes

  • Shell becomes read-only.
  • Files are limited to the agent workspace.
  • No network, browser, web, apps, or notify tools.
  • No skills, and no extra commands or paths.

When to use it

Use secure mode for agents that touch sensitive systems or read untrusted content. It is the profile you hand to a task where you would rather the agent be too limited than too permissive.

Turn it on

  1. Open the agent’s settings (gear icon).
  2. Go to Capabilities.
  3. Toggle Secure mode on.
  4. Save capabilities.

What it cannot do

A secure-mode agent cannot be talked into a wider profile by a prompt. The restriction is enforced outside the model, so no instruction can loosen it.

Combine with budgets

Secure mode limits what the agent can do; a budget limits what it can spend. Together they make a task safe to leave running unattended.