Skip to main content

​
1. Install Operator in your coding agent

Follow the integration steps for the coding agent you use. For Claude Code, open the command input and run: 
/plugin marketplace add operatorlabs/operator-plugins
/plugin install operator
Restart Claude Code once so the plugin, hooks, MCP config, and Skill all load. On the first run in a Claude Code session, the Operator SessionStart hook will start an OAuth flow so you can sign in and store your key under the plugin. Once that succeeds, your sessions are ready to use Operator commands.

​
2. Enable Operator for a project directory

With the plugin installed and authentication set up:
  1. Open a repository in Claude Code.
  2. Start a new session (the SessionStart hook will check your key and status).
  3. Enable Operator for the current directory, for example:
/operator:activate
You can optionally add --max-iterations n to set the max number of stops to trigger; the default is 10. This creates a hidden project config directory, writes a project local config file with max_iterations and other loop settings, and marks the directory as Operator managed. Once enabled, tasks and goals you describe in that directory are tracked across stops.

​
3. Send a task to Operator

If Operator is activated, it will trigger for the next query that you send in Claude Code, keeping the loop active for as long as the original task is not completed. Operator also has a console UI where you can ask what happened on your goals. From the console you can:
  • Ask what your coding agent has done so far.
  • Ask which tasks are finished, in progress, or blocked.
  • Ask why a previous run stopped or what it tried last.

​
Next steps