Works with the AI tools you already use
Claude Code
Cursor
Codex CLI
Muse
Dots+22 moreClaude CodeCursorCodex CLIMuseDots
OpenClaw
Manus
Grok
Grok Bot+18 agentsShared Rate Budget for Multi-Agent Setups
Several agents on one account? One shared request budget they all obey, and a 429 seen by one pauses all of them.
$15
Shared Rate Budget for Multi-Agent Setups
Example session with this skill installed
I run a scout agent and a posting agent on the same X account from cron, plus an analytics job. We got rate limited last night. Set up pacing so it doesn't happen again.
- Read your context and instructions
- Compiled the shared rate budget
Summary
All three jobs now share one budget for the account, start at different minutes, and a 429 seen by any of them pauses all three.
Setup I ran
python3 scripts/budget.py config x-main --limit 30 --window 3600 --min-gap 20
| Job | Schedule (was) | Schedule (now) | Per-run cap |
|---|---|---|---|
| scout | :00 | :05 | 6 searches (take x-main each) |
| poster | :00 | :25 | 4 replies |
| analytics | every hour | :45 every 3 h | 1 scrape |
Added to each job
export AGENT_NAME=scout
if ! python3 scripts/budget.py take x-main; then
echo "budget says wait; ending run"; exit 0
fi
On any 429 / "try again later":
python3 scripts/budget.py hit429 x-main --retry-after 900 --agent "$AGENT_NAME" && exit 0
Current state
x-main: used 11/30 per 3600s | cooldown 0s | strikes 0 | by agent {"scout": 6, "poster": 4, "analytics": 1}
Next steps
- Keep the limit at 30/hour for a clean week before raising it.
- If
statusshows one job using most of the budget, cut that job first. - Do not log in again or switch accounts after a 429; it is a pacing problem, not an auth problem.
Connects securely to your tools. The creator never sees your data.
About this skill
What it prevents
Rate limits belong to the account, not the agent. Three agents that are each "polite" are still one rude account. Per-agent backoff does not help: while agent A waits, B and C keep spending the same budget, and the platform sees one account hammering it. Retries with jitter spread requests out; they do not reduce them.
The incident behind it (true)
A developer ran three agents on one X account: a trend scout running searches, a writer posting replies, an analyst scraping the account's own stats. Each had its own pacing. One evening X returned 429 Too Many Requests on the account's own endpoints (the calls the site makes on every page load), not just on search. No single agent was over anything it knew about. What fixed it was not smarter retries: pause about 2 hours, stagger the jobs (:05 and :25), cap the scout at about 6 searches per run, let the analyst scrape once, stop reloading pages. The next analyst run scraped 143 rows with no 429. This skill is that fix, enforced by a script every agent calls.
What you get
budget.py(stdlib Python, file-locked JSON ledger per account):config x-main --limit 30 --window 3600 --min-gap 20take x-mainbefore every action: exit 0 = go, exit 3 = WAIT with seconds and reasonwait openai-org --cost 5 --max-wait 120to block up to a caphit429 x-main --retry-after 900: shared escalating cooldown (15 min, 30, 60, 120, capped at 4 h; strikes reset after 24 h clean). Every agent on that account waits.status: usage by agent, so you can see which role is eating the account- Tested with 8 concurrent takers on a 5-request budget: exactly 5 go.
- The rules: one bucket per credential, take before you act, a 429 belongs to everyone, escalate instead of jitter, stagger schedules, count the hidden page-load calls of browser automation, start at about 50% of documented limits, and never switch accounts or keys to get around a limit.
- Copy-paste snippets for shell and Python, and the exact per-run instructions for the agent.
Works for social accounts, LLM API keys shared by parallel workers, scrapers, and any API several cron jobs touch.
How to use
Skills on Agensi run inside your AI agent. Copy the prompt, paste it into your agent, and it connects to Agensi and installs this skill for you. If a payment is needed, your agent shows it to you first.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill passes our trust checks below.
Security scanned
Passed our 8-point scan before listing
Fresh listing
Recently published to Agensi
30-day refund
Not a fit? Get your money back
Trust & safety
Security scanned
Verified clean today
- Passed all security checks, Safe to install