codex config forge

// byok profile generator for the openai codex cli

responses api ~/.codex/<profile>.config.toml ⊞ all tools

provider

responses-api only

↪ Need a Chat-Completions-only provider (Mistral, DeepSeek, Anthropic, Gemini…)? Codex speaks the Responses API only — spin up a proxy with the LiteLLM Config Forge, then add it here as a Custom provider.

model

model *

model_reasoning_effort reasoning models only

provider details

provider id [model_providers.<id>]

Azure resource *

→

base_url *

api-version azure query param

env_key name of env var holding the key

Codex reads the key from this env var at runtime — export it in your shell. The key itself is never written to disk.

profile & launch

profile name *

→

reasoning summary

approval_policy

sandbox_mode

--yoloadd to launch command — alias for --dangerously-bypass-approvals-and-sandbox (skips approvals & sandbox).

① provider block · append to ~/.codex/config.toml

~/.codex/config.toml

② profile file

③ launch

~ terminal