Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
Version History
Updated newsletter CTAs with UTM tracking and skill-specific messaging
What This Skill Does
Tracks API request usage in a rolling window and assigns tiers (ok → cautious → throttled → critical → paused). Agents check their current tier before expensive operations and reduce activity accordingly. On a real 429 error, exponential backoff is calculated and recovery is scheduled automatically.
Unlike retry loops or monitoring dashboards, it prevents 429s before they happen by progressively downshifting agent activity, preserving context and avoiding wasted token spend.
When to use it
- Preventing 429 errors in long-running autonomous agent loops
- Throttling sub-agent spawning when approaching Claude Max limits
- Managing rate limits across multiple concurrent agent instances
- Keeping background cron jobs from burning through rolling windows
- Auto-recovering agent workflows after hitting a rate limit wall
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: Run a multi-step research task using sub-agents
- 1Runs `python3 scripts/rate-limiter.py gate` before starting work
- 2Reads exit code: 0 means proceed normally, 1 means reduce activity, 2 means stop non-essential work
- 3Spawns sub-agents and completes the research task
- 4Records token usage with `python3 scripts/rate-limiter.py record 2000`
- 5On 429 error, runs `python3 scripts/rate-limiter.py pause` and waits for auto-resume
Research completes without rate limit interruptions; state file updated with rolling usage data