Skip to main content
This page covers the top-level commands outside the grouped subcommands.

​
Start The Runtime

operator
What it does:
  • loads ~/.operator/operator.yaml
  • loads runtime.env_file if configured
  • installs any missing bundled skills
  • starts the configured transports
  • starts the job runner
  • starts memory workers if enabled
The runtime requires at least one configured transport. A headless-only config cannot start the long-running service.

​
operator init

operator init
Scaffolds:
  • operator.yaml
  • .env
  • SYSTEM.md
  • agents/operator/AGENT.md
  • agents/operator/workspace/
  • jobs/
  • skills/
  • shared/
  • logs/
  • state/
Behavior:
  • safe to re-run
  • existing files are not overwritten
  • installs bundled skills that do not already exist

​
operator setup

operator setup
operator setup --run
Guided onboarding for:
  • provider selection
  • timezone
  • provider API key
  • Slack bot/app tokens
  • first admin username
  • first Slack identity
Useful flags:
FlagMeaning
--provideranthropic, openai, or gemini
--timezoneexplicit IANA timezone
--usernameadmin username to create
--api-keyprovider key to persist
--slack-bot-tokenbot token
--slack-app-tokenapp token
--slack-userSlack user ID
--forcere-prompt secrets even if already present
--runstart the runtime after setup

​
operator config

operator config
Prints the resolved config as JSON after validation and env-file loading. Use this to confirm:
  • agent inheritance
  • runtime timezone
  • role definitions
  • memory settings
  • transport env var names

​
operator agents

operator agents
Shows:
  • configured agent name
  • frontmatter description from AGENT.md, when present
  • transport type
  • resolved models
  • whether AGENT.md and workspace/ exist
This is the quickest way to confirm that an agent is both configured and scaffolded on disk.

​
operator tools

operator tools
Lists the registered built-in tools. Important caveat: this command does not include transport tools like Slack’s list_channels, read_channel, or read_thread. Those are added only during live runtime execution.

​
operator logs

operator logs
operator logs -n 200
operator logs -f
Tails ~/.operator/logs/operator.log. Options:
OptionMeaning
-n, --linesnumber of lines to show, default 50
-f, --followfollow new output
If the log file does not exist yet, the command exits with an error.

​
Module Entrypoint

If the operator script is not on your PATH, the package also supports: 
python3 -m operator_ai
python3 -m operator_ai setup --run