CLI integrationsCodex
Add a custom provider in ~/.codex/config.toml, point base_url at Framewharf and use wire_api responses.
Declare it in ~/.codex/config.toml. A .codex/config.toml file inside a repository has its model_provider and model_providers — keys ignored, so don't put your API key there. The provider id must not be openai, ollama or lmstudio.
Set model to any id in the catalog. To run once with a different model:
If Codex opens the ChatGPT login screen, exit and check model_provider, base_url and FRAMEWHARF_API_KEY.
Paste the exact id from GET /v1/models. A few ids available right now:
curl -fsSL https://chatgpt.com/codex/install.sh | sh
codex --versionexport FRAMEWHARF_API_KEY="sk-your_api_key"model_provider = "framewharf"
model = "gpt-6.1-sol"
[model_providers.framewharf]
name = "Framewharf"
base_url = "https://api.framewharf.com/v1"
env_key = "FRAMEWHARF_API_KEY"
wire_api = "responses"codex -m claude-haiku-5-5 "Explain this repository"cd ~/your-project
codex
# a single prompt, without the TUI
codex exec "Explain this repository"| Model id | Details | Copy |
|---|---|---|
claude-haiku-5-5 | Claude Haiku 5 5 | |
gpt-6.1-sol | Gpt 6.1 Sol | |
gemini-4-hydrogen | Gemini 4 Hydrogen | |
kimi-k3 | Kimi K3 | |
deepseek-v4.1-flash | Deepseek V4.1 Flash | |
qwen4-max | Qwen4 Max |