openclaw-plugins

Appearance

native-scheduler ​

The native-scheduler plugin lets OpenClaw agents manage OS-level scheduled jobs — launchd on macOS, systemd timers on Linux, cron as a fallback, and Windows Task Scheduler on Windows.

Why native scheduling? ​

OpenClaw's built-in cron system runs tasks through the agent loop, consuming LLM tokens on every tick. For deterministic scripts (health checks, notifications, data collection), that's wasteful. native-scheduler runs scripts directly through the OS scheduler and only invokes the agent when the script's output demands it.

How it works ​

  1. The agent calls the native_scheduler tool with an upsert action and a job definition.
  2. The plugin materializes a wrapper runner script that handles stdin/stdout contract enforcement, health tracking, and failure callbacks.
  3. The wrapper + job are registered with the platform's native scheduler.
  4. On each trigger, the wrapper pipes a NativeSchedulerRunContext to the script's stdin and parses the script's stdout as a NativeSchedulerResult.
  5. Based on the result type (noop, prompt, or message), the wrapper either does nothing, triggers an agent session, or delivers a message directly to a channel — all without burning tokens for the decision logic.

Key features ​

  • Cross-platform: launchd, systemd, cron, Windows Task Scheduler
  • Typed script contract: Scripts receive structured input and return structured output via @spectratools/native-scheduler-types
  • Zero-token message delivery: message results go directly to channels without an LLM turn
  • Health tracking: Per-job health files with success/failure counts, streaks, and timestamps
  • Failure callbacks: Configurable actions on script crash or timeout (run a command or fire an OpenClaw event)
  • 12 tool actions: Full lifecycle management from the agent

Next steps ​