Codex CLI

Configure Codex with a user-level model provider that speaks the Responses API.

Configuration

model = "gpt-5.6-sol"
model_provider = "llmtap"

[model_providers.llmtap]
name = "llmtap"
base_url = "https://www.llmtap.dev/api/v1"
env_key = "LLMTAP_API_KEY"
wire_api = "responses"

export LLMTAP_API_KEY='sk-live-your-key'

Errors

Codex provider settings belong in ~/.codex/config.toml. Keep the API key in an environment variable and verify the selected model is available in the llmtap model list.