Video summary

FREE & UNLIMITED Codex with GPT-5.5 Real Method #ClaudeAI #Codex #ClaudeCode

Main summary

Key takeaways

Technology

Overview

The video demonstrates a “second method” for making Codex work with a different backend/LLM—specifically “LLM 7” (and it also mentions switching to GPT-5.5).

The speaker claims Codex initially fails to use the desired model(s) (referred to as “LLM 7”) and instead hits a Codex message/usage limit. The workaround reconfigures Codex so additional models and reasoning options can be used without being blocked by the original limit.


Key claims / observations

  • Codex (Microsoft Store install) initially shows an error such as:

    • “out of Codex messages / usage limit reached” when sending messages.
  • Codex includes a “reasoning level” indicator, described as limited at first.

  • To enable more capabilities, the speaker:
    • Fully closes Codex (must exit from the taskbar, not just the window).
    • Reconfigures Codex using a local Python proxy server that forwards requests.
  • After the proxy setup, Codex reportedly shows:
    • More reasoning models/capabilities, including options like “Extra high” and “Max”
    • Custom model selection
    • The ability to start a new chat (previous chats may not function with the new backend)

Tutorial / setup workflow (step-by-step)

  1. Install Codex from the Microsoft Store.
  2. Sign in with ChatGPT to open Codex and verify the initial state:
    • Confirm the credit/message usage limit prevents chatting.
  3. Close Codex completely:
    • Exit from the taskbar (not just closing the window).
  4. Download a Python project (repo) as a ZIP and extract it.
  5. In the extracted project:
    • Open the project in a terminal
    • Install dependencies using the command shown in the project requirements.txt / referenced from the README
  6. Update Codex configuration:
    • Locate config.toml in the Codex folder under your user directory (example path shown):
      • C:\Users\<your user>\...
    • Replace its contents using the values from the project README
    • Set the model (initially referenced as “default”) and save
  7. Configure environment variables for the backend model:
    • Generate an LLM 7 API key from an “LLM 7 website” (login → create key → set expiry; example uses expiry in 2030)
    • Copy the API key into the terminal command from the README to set an environment variable (example pattern like):
      • set LLM API key <your key>
  8. Run the proxy server:
    • Execute a Python command (the README includes variants for Linux/macOS)
    • Verify the server is running (e.g., listening on a port)
  9. Validate in Codex:

    • Reopen Codex
    • Confirm requests are going to the local proxy server (a request line appears)
    • Start a new chat

      Warning: do not reuse the previous chat

    • Test with “hi” and verify claims:

      • No usage limit appears now
      • Extra high reasoning / Max style options become available 10. Model switching test:
    • Change Codex model to GPT-5.5 and set medium thinking
    • Verify it works

Support / troubleshooting notes

If code/server issues occur, the speaker suggests viewers can:

  • Leave a comment, or
  • Contact them via Telegram (or another “proper platform” referenced in the video)

Main speakers / sources

  • Primary speaker: an unnamed individual addressing “guys” (no name provided in the subtitles)

Sources referenced

  • Microsoft Store (Codex installation)
  • ChatGPT (used to sign into Codex)
  • “LLM 7 website” (used to generate an API key)
  • A GitHub/repo-style Python project (downloaded as ZIP; README contains commands and config details)

Original video