On Friday, October 9, my blog agent needed eight images for one post. It’s a fast agent, so it asked for all eight at once. Three came back rejected with HTTP 429, “too many requests.” The other five were fine. The agent retried the three one at a time, and all three worked.
Later that morning, my social agent asked for five images in parallel. One got a 429, and the script it was running treated that as a hard failure and quit.
The next day I changed one thing: send them in batches of three, and wait eight seconds before starting each new request. Across the next two image-heavy runs that was 12 images, 12 successes, and zero retries. That’s today’s tip.
Why AI agent rate limits hit when the work is independent
Back in July I wrote that parallel agents win when the work is independent, and I used “seven section images, done in the time of the slowest render” as the example. The work really is independent. Each image has its own prompt. But all of them go through the same API key, and the key has a speed limit. Independent work doesn’t come with independent rate limits.
Two things made this worse than it looked:
⚡ GET THE AI EDGE
Weekly AI tips that actually save you time and money. No fluff, no hype — just what works.
- The key is shared. I run about ten brands, and they all use one image account. When I checked the account during an earlier failure, it showed renders I hadn’t started. A sibling brand was bursting at the same minute. Your agent isn’t the only one spending your limit. Your other automations, your teammates and your scheduled jobs all count too.
- The health check lied politely. The account endpoint said 2,999 requests remaining. That’s the limit for reading data. Starting a new image job has its own limit, it’s much tighter, and the header doesn’t report it. “Plenty of headroom” was true, just not for the thing I was doing.
The fix: batch, stagger, retry the one that failed
- Cap the batch. I use three. Five at once has worked on some days and lost one on others. Eight at once lost three. Start small and only raise it once you have a few clean runs.
- Stagger the starts. Wait a few seconds between requests, even inside a batch. Eight seconds costs me under a minute on a seven-image post. A failed render costs more than that just to notice.
- Retry only what failed, and retry it alone. A 429 means “slow down,” not “broken.” Wait, then send the one rejected request by itself. Don’t re-run the whole batch, because you’ll pay again for the ones that already worked.
- Don’t let a 429 kill the run. If your script quits on the first rejection, one busy minute costs you the whole job. Make sure the code that calls the API waits and retries before it gives up.
One caution: a 429 can also mean you’ve used up your monthly quota, and waiting won’t fix that. When my image compressor hit its monthly cap in September, it kept answering 429 for days. Before you retry, read the error text. “Too many requests” means wait a few seconds. “Monthly limit exceeded” means you need a bigger plan or a fallback, and I covered that case in what happens when one of your agent’s tools breaks.
Where to look in your own setup
Search your agent’s instructions and scripts for anything that loops over a list and calls an API: images, emails, scrapes, CRM updates. Then ask two questions. How many requests start at the same moment? And what happens to the run when one of them gets a 429? If the answers are “all of them” and “it stops,” you’ve found the next failure before it finds you.
The takeaway: parallel work still shares one speed limit. Batch in small groups, space out the starts, and retry only the request that failed. My image steps went from 3 failures in 8 to 12 for 12 without spending anything extra.
Not sure which of your automations are racing each other for the same key? Book an automation strategy session and we’ll map it together.
📥 FREE: THE AI PLAYBOOK
The exact tools and workflows I use to run a one-person agency. 25 years of marketing experience distilled into an actionable guide. Yours free.
AI & Digital Marketing Strategist
With over 25 years in digital marketing, e-commerce, sales, and international business, Jon Jones is a seasoned entrepreneur dedicated to empowering small businesses through innovative AI solutions. His journey began in investment banking on Wall Street, guiding companies through capital acquisition and complex markets, and has expanded to include expertise in web development, brand design, product development, and courses and training.
Fluent in Mandarin and with a rich foundation in both finance and tech, Jon launched JonJones.ai to provide accessible, transformative tools for small businesses to streamline operations, generate leads, and build sustainable growth. His mastery of AI, automation, and digital strategy has enabled numerous businesses to bridge the gap between traditional practices and modern technology, ensuring they thrive in today’s digital landscape.
