Sign in with ChatGPT (Pro): “You’ve hit your usage limit” after a single ~100s analysis on gpt-5; stops reproducing after disabling MCP servers (Codex CLI 0.23.0)
Summary
After updating to Codex CLI 0.23.0, my very first long analysis (\~100 seconds) using gpt-5 (Reasoning Effort: Medium) ended with:
🖐 You've hit your usage limit. Limits reset every 5h and every week.
I was authenticated via Sign in with ChatGPT (Pro). I had several MCP servers configured (mainly Context7 and Memory MCP). After disabling all MCP servers, I’ve been able to run at least three similar analyses without hitting the limit.
Context: the README documents Sign in with ChatGPT (Plus/Pro/Team) and shows that you can use the latest models including gpt-5.([GitHub][1])
Environment
- Codex CLI: 0.23.0
- OS: Ubuntu 24.04 (VS Code terminal)
- Auth: Sign in with ChatGPT (Pro)
- Model:
gpt-5 - Reasoning Effort: Medium
- MCP: previously enabled (Context7, Memory MCP), now disabled
My steps (original failure)
- Launch
codexand authenticate via Sign in with ChatGPT. ([GitHub][1]) - Ask for a repo-wide analysis (e.g., “Check
REFACTORING_PLAN.mdand verify Stage A”). - Let it run \~100 seconds.
- Observe the usage-limit message above.
Actual
- After a single long run, the TUI shows usage-limit.
- I don’t have the old logs; total consumption at that moment was well under \~400k tokens.
- After disabling MCP, I haven’t reproduced the issue (at least three similar runs).
Ask to the maintainers
- Please clarify what limits apply to ChatGPT Pro when using Sign in with ChatGPT in Codex CLI: exact caps (time windows, per-model considerations), how usage is counted (including any MCP contributions), and where to view remaining quota for this auth mode.
I looked through the README and Help Center but couldn’t find a definitive page with limit numbers or a quota dashboard — I may have missed it; a pointer would be appreciated. ([GitHub][1], [OpenAI Help Center][3])
- Also, did 0.23.0 introduce any change in quota/usage handling for ChatGPT-based auth? On the previous version I could use MCP with the same settings without hitting this limit. ([GitHub][2])
[1]: https://github.com/openai/codex "GitHub - openai/codex: Lightweight coding agent that runs in your terminal"
[2]: https://github.com/openai/codex/releases?utm_source=chatgpt.com "Releases · openai/codex"
[3]: https://help.openai.com/en/articles/11381614-codex-cli-and-sign-in-with-chatgpt "Codex CLI and Sign in with ChatGPT | OpenAI Help Center"
This issue has 2 comments on GitHub. Read the full discussion on GitHub ↗