Introducing Conductor Cloud →Skip to docs content
Conductor

Codex

How Conductor runs Codex and how Codex usage is billed

Conductor runs Codex through the Codex setup available to Conductor. That can be Conductor's bundled Codex binary, your system Codex binary if configured, and the OpenAI or Codex authentication available in Conductor.

OpenAI's coding agent running through the OpenAI/Codex setup available to Conductor.

Setup
OpenAI or Codex sign-in, API-key configuration, model access, and local Codex settings.
Billing
OpenAI account, subscription, ChatGPT credits, API key, model access, and approvals.
Best fit
OpenAI agent workflows, reasoning controls, personalities, implementation, debugging, and review.
Controls
Plan Mode, Fast Mode, reasoning controls, personalities, checkpoints, and skills where supported.

Models

Use the provider docs for the live model list and pricing.

ModelGuidance
GPT-5.5

Recommended starting point for most Codex tasks.

OpenAI describes it as the strongest option for complex coding and research workflows.

GPT-5.4

Frontier option for professional coding and tool-use workflows.

Use when your account and Codex setup expose it.

GPT-5.4 mini

Faster, lower-cost option for lighter coding tasks and subagents.

Good when latency and cost matter more than maximum capability.

GPT-5.3 Codex Spark

Research preview option for near-instant coding iteration.

OpenAI documents gpt-5.3-codex-spark as available to ChatGPT Pro users.

Capabilities and controls

These notes describe how Conductor wraps the harness and where provider controls still apply.

Model guidance is upstream

Conductor shows the models your Codex setup can use; OpenAI owns model availability and access.

Approvals still matter

Codex approval and permission settings control what the agent can do inside a session.

API-key usage is token billed

OpenAI documents API-key Codex usage as paid based on the tokens Codex uses.

When to use Codex

Use Codex when you want OpenAI's coding agent inside a Conductor workspace. It fits implementation, debugging, review, and focused coding tasks that should stay attached to a branch, terminal, diff, checks, and pull request path.

Codex can use Conductor session controls such as Plan Mode, Fast Mode, reasoning controls, personalities, checkpoints, and skills when those controls are supported by the selected model and Codex setup. For the full control matrix, see Agent modes.

Setup

Conductor uses the OpenAI or Codex authentication available to it. If Codex cannot start, check your OpenAI authentication, subscription setup, API key configuration, and selected model.

You can sign in with the Codex CLI outside Conductor, or configure API-key based access where your workflow requires it. For provider-level setup in Conductor, see Configure model providers.

Models

OpenAI's Codex model docs currently recommend starting with gpt-5.5 for most Codex tasks. They recommend gpt-5.4-mini when you want a faster, lower-cost option for lighter coding tasks or subagents.

Treat provider model guidance as upstream guidance, not a Conductor guarantee. The models visible in Conductor depend on the Codex setup, account, subscription, API access, and model availability at the time you start the session.

Usage and billing

Conductor does not bill or meter Codex usage. Codex usage is controlled by your OpenAI account, subscription, credits, API key, model access, local resources, and configured approvals.

OpenAI's Codex pricing docs describe credit-based usage and API token-based rates for Codex models and features. Usage limits vary by plan, model, task size, local or cloud surface, and speed settings. Check OpenAI's live pricing page for current rates and limits before planning heavy parallel work.

Provider docs

On this page