# An agent stuck in a retry loop overnight turns per-use pricing into a surprise bill. What limits do you set? — retained record

> Retained reddit record from reddit:AI_Agents.

Canonical: https://highsignal.app/data/records/ed0b2c2c61d17737

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:AI_Agents)  published 2026-11-17 13:45 ingested 2026-11-18 02:36

[original source ↗](https://www.reddit.com/r/AI_Agents/comments/1xyktfy/an_agent_stuck_in_a_retry_loop_overnight_turns/)

## retained text



```
Scenario I think about a lot: an agent runs on a pay-per-use model API, hits an error it can't get past, and keeps retrying all night. Nobody notices until the invoice or the usage dashboard.

Full disclosure: I built a rankings site that scores agent tools, including model APIs. Of the ones I track, OpenAI Realtime API scores 8.75/10, OpenAI API 8.45, xAI Grok API 8.15 and Anthropic Claude API 8.1, and all of them bill per use. But I'm more interested in how people actually control spend.

What I'm curious about:
- Do you rely on the provider's account limits, or do you cap spend and retries inside the agent itself?
- Is your provider limit a hard stop or just an alert, and who gets that alert?
- Has a loop or retry storm ever actually cost you, and what did you change afterward?

(link in my profile if anyone wants it)
```



document fetched 2026-11-18 00:19 · canonical https://reddit.com/r/AI_Agents/comments/1xyktfy/an_agent_stuck_in_a_retry_loop_overnight_turns
