Rate limit errors across LLM providers

How rate limit errors show up across Gemini, OpenAI, Anthropic, OpenRouter and more, and how to route around them.

Rate limits are the one error class you should expect in healthy operation: providers cap requests and tokens per minute, and bursty agent loops hit those caps fast. Respect Retry-After, add jitter to retries, and spread load across providers instead of hammering one. Routing with automatic fallback turns a 429 into a request served by the next provider.