Skip to content
BrainRoad BrainRoad
Documentation Menu

Console Overview

A real terminal inside your AI helper's private workspace, for power users who want to look under the hood.

On this page

The Console gives you an interactive shell inside your AI helper’s private workspace. It is a real terminal session, not a simulation: you type commands, they run inside the helper’s isolated environment, and you see the output live.

You will not need the Console for day-to-day work. The dashboard covers reviewing drafts in the Inbox, managing your Business Brain, and starting or stopping your helper. The Console is for power users who want to inspect files, debug, or run developer tools directly.

How to Open It

The Console lives under Console in the dashboard sidebar (in the advanced section). It needs your helper to be running; if the helper is stopped, start it from the Helper page first. It works for both helper runtimes, OpenClaw and Hermes.

The session runs inside tmux, so it survives brief connection drops. If the connection does close, reopen the Console page to start a fresh session. Disconnecting never affects your helper.

What You Can Do

  • Run standard shell commands (ls, cat, ps, top, df -h)
  • Inspect the helper’s files and logs
  • Install extra packages or scripts for your own projects
  • Use the preinstalled coding tools: Claude Code CLI and Codex CLI

If your helper runs OpenClaw, the openclaw command-line tool is also available. Two commands are worth knowing:

openclaw status    # health overview
openclaw doctor    # diagnose configuration issues

For the full OpenClaw CLI and configuration reference, use OpenClaw’s own documentation at docs.openclaw.ai. Runtime settings, channels, skills, and logs are best managed in the runtime’s own screens, opened from your dashboard, rather than by editing files in the shell.

What the Console Does Not Change

The review step still applies. Nothing customer-facing is sent from the shell: outbound email from your helper’s managed mailbox only goes out through drafts you approve in the Inbox, and the workspace’s network rules enforce that boundary. The Console is a window into the workspace, not a way around the review process.

One practical caution: BrainRoad manages parts of the helper’s runtime configuration (networking, authentication, the connection to your Business Brain) each time the workspace starts. Hand-editing configuration files can put your helper in a broken state until those managed settings are restored. Prefer the dashboard and the runtime’s own settings screens; if you do experiment and something breaks, openclaw doctor is the first thing to try, and stopping and starting the helper from the Helper page gives you a clean restart.