Feature Request: Surface rate limit errors (429) to assistant for self-throttling

Resolved 💬 2 comments Opened Jan 14, 2026 by M2Scott Closed Feb 26, 2026

Problem

When Claude Code hits API rate limits (429 errors), the assistant is not notified and continues making requests until the process crashes with exit code 1. The rate limit errors are handled internally by the extension but never surfaced to the assistant.

Current Behavior

  1. Assistant makes multiple tool calls (especially in parallel or with agents)
  2. Extension handles API requests and hits 429 rate limit
  3. Extension retries but continues hitting rate limits
  4. Process crashes with "Claude Code process exited with code 1"
  5. Assistant never sees the 429 errors and cannot adjust behavior

Requested Feature

Pass rate limit errors back to the assistant as part of tool results or system messages, so the assistant can:

  • Detect when it's hitting rate limits
  • Self-throttle by reducing parallel requests
  • Wait between tool calls when rate limited
  • Provide better user feedback about what's happening

Benefits

  • Fewer crashes from rate limiting
  • More graceful degradation when limits are hit
  • Assistant can adapt its request patterns in real-time
  • Better user experience with transparent communication about rate limits

Example Implementation

When a 429 error is encountered, return it as a system message:

<system-reminder>Rate limit encountered (429). Consider reducing request frequency.</system-reminder>

Or include it in tool results so the assistant knows to slow down.

This would enable the assistant to self-regulate and avoid crashing the entire process.

View original on GitHub ↗

This issue has 2 comments on GitHub. Read the full discussion on GitHub ↗