MegaBrain University · guide

Codex CLI Custom Provider: Run Codex on GPT-6 and Other Models With One API Key

Updated November 11, 2026 · 5 min setup

What is Codex

Codex is OpenAI’s open-source coding agent. It runs in the terminal (codex), in IDEs and headless in CI (codex exec): it reads the repo, follows the project’s AGENTS.md, edits files and runs commands inside a sandbox with approval modes, and uses MCP servers.

Codex signs in with ChatGPT by default, but its config accepts custom model providers. MegaBrain serves the OpenAI Responses API that Codex speaks, so one key and one balance give Codex GPT-6.1 Sol and Terra, and the same key works for every other model and agent in MegaBrain University.

Setup, step by step

1. Step 1. Get a MegaBrain API key

   Sign up with GitHub, Google or email, add funds on the Balance page and create a key on the API key page. The key works for every text, image and video model, and each request is paid from your balance — no subscription.

   Put the key in an environment variable so it never lands in a config file you share:

   # ~/.zshrc or ~/.bashrc
   export MEGABRAIN_API_KEY="your-megabrain-key"

   Copy

   curl https://getmegabrain.com/api/gateway/v1/chat/completions \
     -H "Authorization: Bearer $MEGABRAIN_API_KEY" \
     -H "Content-Type: application/json" \
     -d '{"model": "anthropic/claude-sonnet-5.5", "messages": [{"role": "user", "content": "Say OK"}]}'

   Copy

   Check the key before you touch the agent. A JSON reply with choices means the key, the balance and the model all work:

2. Step 2. Install Codex

   With npm, or with Homebrew on macOS:

   npm install -g @openai/codex
   # or
   brew install --cask codex

   Copy

3. Step 3. Add MegaBrain as a model provider

   Edit ~/.codex/config.toml. The provider must live in this user-level file: a project’s .codex/config.toml cannot set model_provider.

   model = "openai/gpt-6.1-sol"
   model_provider = "megabrain"
   
   [model_providers.megabrain]
   name = "MegaBrain"
   base_url = "https://getmegabrain.com/api/gateway/v1"
   env_key = "MEGABRAIN_API_KEY"
   wire_api = "responses"

   Copy

   env_key names the environment variable that holds the key, so the key itself never sits in the file. wire_api = "responses" makes Codex call /responses, which MegaBrain serves.

4. Step 4. Run Codex and check

   Open a repo and run codex. Ask for something that touches files — “add a unit test for utils/date.ts and run it”. If Codex edits the file and runs the test, the provider and tool calls work. For one task without the UI: codex exec "summarize what this repo does".

5. Step 5. Switch models with profiles

   A profile is a file next to the config with its own top-level keys. Create ~/.codex/cheap.config.toml and run codex --profile cheap for routine work (Codex 0.134 and later; older versions used [profiles.<name>] tables):

   # ~/.codex/cheap.config.toml
   model = "openai/gpt-6.1-terra"
   model_reasoning_effort = "low"

   Copy

Which model to choose

Codex is tuned for OpenAI models: use openai/gpt-6.1-sol by default and openai/gpt-6.1-terra for cheaper routine runs. Other models in the catalog are served over the same Responses endpoint, but Codex’s prompts and tools are built around GPT; try another model on a small task first, and for Claude prefer Claude Code.

Every model with its price and context window is in the catalog. They all work with the same key.

What you can do

Business use cases

The CI job failed with the log in ci.log. Find the root cause, fix it with the smallest change, run the failing tests until they pass, and write a 3-line summary of the cause and the fix to FIX.md.

Troubleshooting

FAQ