Claude Code rate limits operate as three independent constraints: requests per minute (RPM), tokens per minute (TPM), and daily token quota. The dashboard percentage only reflects the daily quota, which is why developers can see 6% usage and still hit a 429 error. Claude Code consumes 10-100x more tokens than chat due to multi-turn conversations, growing context windows, and tool-use round trips. The guide covers all tier limits (Free through Tier 4, Pro, Max, Enterprise), how to decode 429 vs 529 errors, reading rate limit response headers, and practical mitigation strategies including .claudeignore files, exponential backoff with jitter, session compaction, model routing, and building a team-level rate limit monitoring proxy.
Table of contents
Table of ContentsThe 6% MysteryHow Claude Code Consumes Your API QuotaThe Three Rate Limit Types ExplainedRate Limit Tiers: Free, Pro, Max, and Enterprise ComparedDecoding the Error MessagesDiagnosing Which Limit You're Actually HittingPractical Strategies to Avoid Rate Limits in Claude CodeBuilding a Rate Limit Monitor for Your TeamKey Takeaways and Quick ReferenceSort: