Codex

Codex needs a provider that speaks the OpenAI Responses API — it dropped Chat Completions in February 2026. We serve Responses, including the encrypted reasoning it expects back on the next turn.

The short way

npx turboprivate@latest login
npx turboprivate@latest codex

Your existing Codex config is kept — the provider block is added to it, and a plain codex behaves exactly as before.

The manual way

# ~/.codex/config.toml — USER level only. Codex ignores model_provider in a
# project-local .codex/config.toml and prints a warning.
[model_providers.turboprivate]
name = "TurboPrivate"
base_url = "https://api.turboprivate.ai/v1"
env_key = "TURBOPRIVATE_API_KEY"
wire_api = "responses"

export TURBOPRIVATE_API_KEY=tp_live_…
codex -c model_provider=turboprivate -c model=ds-flash