Integrations

Codex by CCS

Configure Codex with UnifyLLM through CC Switch.

Integration Overview

This page explains how to connect Codex to UnifyLLM through CC Switch. Add UnifyLLM under Unified Providers, choose the NewAPI preset, enable Codex, use the /v1 endpoint, and select the native Responses format.

Install Codex: Desktop or CLI

You can use the Codex desktop app or Codex CLI. Use the official OpenAI Codex page for the desktop product, or install the CLI in a terminal. Complete the official sign-in before adding UnifyLLM in CC Switch.

  1. 1

    macOS / Linux: install the CLI

    Run the official standalone installer from Terminal, iTerm2, or another bash/zsh shell.

  2. 2

    Windows: install the CLI

    Run the PowerShell installer from PowerShell or Windows Terminal. Do not paste PowerShell syntax directly into Git Bash.

  3. 3

    Launch and sign in

    Open a project directory, run codex, and choose Sign in with ChatGPT or another method offered by your current release.

  4. 4

    Verify the command

    Run codex --version. If the command is not found, restart the terminal and check the PATH written by npm or the standalone installer.

bash
1# macOS / Linux2curl -fsSL https://chatgpt.com/codex/install.sh | sh3 4# Windows PowerShell5powershell -ExecutionPolicy ByPass -c "irm https://chatgpt.com/codex/install.ps1 | iex"6 7# cross-platform npm fallback8npm install -g @openai/codex9 10# start in a project11codex

Before You Start

  • API Base URL: https://api.unifyllm.top/v1
  • An active UnifyLLM API key
  • A currently available model ID, for example gpt-5.6-sol. Models, groups, and prices can change; use the live pricing page and console model list as the source of truth.

Configure It in CC Switch

  1. 1

    Install and open CC Switch

    Open provider management and select Codex in the application switcher.

  2. 2

    Open Unified Providers

    Click + and add UnifyLLM under Unified Providers, not under an application-specific provider list.

  3. 3

    Choose the NewAPI preset

    Set the preset type to NewAPI and name the provider UnifyLLM.

  4. 4

    Choose the format and endpoint

    Select Responses (native format), use https://api.unifyllm.top/v1, and paste your API key. Never put the key in screenshots, repositories, or public logs.

  5. 5

    Add and select a model

    Add gpt-5.6-sol or choose the exact ID from Get Models. If the list is empty, check the endpoint, key, and whether /v1 was entered twice.

  6. 6

    Enable the provider

    Enable the card and restart the tool as required by Codex. Add UnifyLLM under Unified Providers, choose the NewAPI preset, enable Codex, use the /v1 endpoint, and select the native Responses format.

Manual Configuration Reference

If you are not using CC Switch, the following is a reference for the common configuration format of this tool. Field names can change between releases, so prefer the tool's own documentation when it differs.

toml
1model = "gpt-5.6-sol"2model_provider = "unifyllm"3 4[model_providers.unifyllm]5name = "UnifyLLM"6base_url = "https://api.unifyllm.top/v1"7env_key = "UNIFYLLM_API_KEY"8wire_api = "responses"

Test and Troubleshoot

After enabling the provider, restart Codex and send a small coding request. Check the console usage log for the selected model and token cost.

API references: Chat Completions, Responses, Messages, and Generate Content.

  • Start with a short request before running a large workload.
  • Match the model ID exactly with the pricing page and console, including any group prefix.
  • Review console usage logs for delivery, the actual model, token usage, and errors.
  • If the tool appends /v1 automatically and you entered /v1 manually, remove the duplicate path.

Official CC Switch References

This guide references the official CC Switch introduction, Add Provider, and Edit Provider. Field names and paths may change after a CC Switch update; use its latest documentation and application UI as the final reference.

Security Reminder

An API key is an account credential. Use a dedicated key, limit where it is stored, and disable/recreate it immediately if it leaks. Never place it in browser code, a public repository, screenshots, or third-party sync services.