Agent Model Connector

Fix Rate Limit / Insufficient Quota in Claude Code and Codex CLI

Step-by-step fixes for the "Rate Limit / Insufficient Quota" error when connecting Claude Code or Codex CLI to DeepSeek, OpenRouter, and other providers.

Error output

429 Too Many Requests {"error":{"message":"Rate limit reached for requests"}}
Common variations
402 Payment Required {"error":{"message":"Insufficient Balance"}}
anthropic: Request failed with status 529: {"error":{"message":"Overloaded"}}

The exact text varies by agent version and provider. Match the status code and the message above.

Quick diagnosis

The status code tells you which layer is throttling you: 429 = rate limit, 402 = balance, 529 = provider overload.

Claude Code 
claude /status
Codex CLI 
codex exec "hi"

Common causes

  • The provider limits requests per minute and the limit window has not reset.
  • The account balance ran out; some providers surface this as 429 or 402.
  • Free-tier request caps are exhausted (OpenRouter free models).
  • Long agentic sessions burst many requests in a short window.

Fix by scenario

1 Happens during heavy agentic sessions

  • Reduce parallelism: fewer subagents or concurrent requests.
  • Switch long-running work to a cheaper/faster model.
  • Add small delays between retries instead of hammering the API.

2 Free tier exhausted

  • OpenRouter free models cap around 20 req/min and 200 req/day per model.
  • Rotating across multiple free models extends the runway.
  • Adding $10 in credits raises the daily free limit.

3 Balance-related rejection

  • Top up the account; some providers report insufficient balance as 429 or 402.
  • Set a budget cap so a runaway session cannot empty the balance.
  • Check the usage page for the rejected request before changing config.

Agent-specific fixes

Claude Code

Fix
Wait for the limit window, top up the account, or switch to a cheaper model. In long sessions, lower CLAUDE_CODE_SUBAGENT_MODEL to a cheap model for subagent calls.
Verify
Retry `claude /status` and a short prompt after the window resets.

Codex CLI

Fix
Reduce parallelism (fewer subagents), switch to a faster model, or raise limits by adding credits.
Verify
Confirm the rejected request in the provider dashboard, then retry `codex exec "hi"`.

Claude Code vs Codex CLI at a glance

Claude Code Codex CLI
Status codes 429 / 402 / 529 429 / 402 / 529
Balance check Provider dashboard Provider dashboard
Cheap model for subagents CLAUDE_CODE_SUBAGENT_MODEL model in config.toml

Prevent it from happening again

  • Set budget caps on the provider dashboard.
  • Prefer cheaper models for routine tasks and keep frontier models for hard problems.
  • Add retry-with-backoff to your workflow instead of instant retries.

Related errors

Timeout / Request Timed Out 401 Unauthorized

Related setup guides

Claude Code with DeepSeek Open the setup guide Claude Code with OpenRouter Open the setup guide Codex CLI with DeepSeek Open the setup guide Codex CLI with OpenRouter Open the setup guide

Frequently asked questions

What are the OpenRouter free-tier limits?

Free models cap around 20 requests per minute and 200 requests per day per model. Adding $10 in credits raises the daily free limit.

Is 529 a rate limit?

529 means the provider is overloaded, not that your key is limited. Wait and retry; switching models or endpoints often helps.

All error guides