Agent Model Connector

Fix Timeout / Request Timed Out in Claude Code and Codex CLI

Step-by-step fixes for the "Timeout / Request Timed Out" error when connecting Claude Code or Codex CLI to DeepSeek, OpenRouter, and other providers.

Error output

Error: Request timed out after 600000ms. Anthropic API Error: 529 Overloaded
Common variations
codex: request timed out after 10 minutes
Error: TimeoutError: Request timed out at Object.fetch (node:internal/…)

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

Quick diagnosis

A timeout right after a small prompt points at the network; a timeout after a huge prompt points at the model.

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

Common causes

  • The model is slow (long reasoning, large context) and the client gave up.
  • A very long prompt or huge context extends time to first token.
  • Restricted or congested network to the provider region.
  • The provider is overloaded (529) during peak hours.

Fix by scenario

1 Times out on huge prompts or repos

  • Shorten the prompt or narrow the context (smaller repo scope, targeted files).
  • Split long tasks into smaller steps instead of one giant request.
  • Use a model with a larger context window or higher speed for bulk work.

2 Times out on small prompts too

  • This points at the network path — run the curl test from the same shell.
  • Check if the provider is overloaded (529) before changing config.
  • Retry during off-peak hours to confirm it is load-related.

3 Timeouts after a config change

  • A new model may reason much longer — lower the reasoning effort or switch models.
  • A proxy or VPN change can add latency — re-run the curl timing test.
  • Compare against the previous working model before assuming the endpoint is at fault.

Agent-specific fixes

Claude Code

Fix
Retry the request, shorten the prompt or context, and switch to deepseek-v4-flash for long sessions. Region-based issues may need a different network.
Verify
Re-run the request when the network is idle and confirm it completes.

Codex CLI

Fix
Reduce context (smaller repo scope), lower reasoning effort for simple tasks, or switch to a faster model. Check for provider overload (529) before changing config.
Verify
Run `codex exec "hi"` with a short prompt to isolate the network path from the workload.

Claude Code vs Codex CLI at a glance

Claude Code Codex CLI
Small-prompt timeout Network path likely Network path likely
Large-prompt timeout Model speed/context Model speed/context
Overload signal 529 Overloaded 529 Overloaded

Prevent it from happening again

  • Prefer faster models for routine work and keep slow reasoning models for hard problems.
  • Keep prompts and context lean; big context is the most common timeout trigger.
  • Run long sessions off-peak or with retry logic in place.

Related errors

Rate Limit / Insufficient Quota Network Error / Connection Failed

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

Can I increase the timeout?

Most clients use a fixed client-side timeout. The practical fixes are a faster model, a shorter prompt, or a better network path to the provider.

Timeout and 529 at the same time — what does that mean?

529 means the provider is overloaded; the timeout is the client giving up while waiting. Wait, retry off-peak, or switch to a less loaded model or endpoint.

All error guides