Concepts

ChatGPT (Subscription)

Run AI nodes on your ChatGPT Plus or Pro plan — no API key, through the Codex CLI you already have.

How it works

Watchflows calls the codex binary on your Mac instead of an HTTP API. You sign in to Codex once with your ChatGPT account, and the app uses that session. Everything an AI node can do works here: plain prompts, JSON-schema structured output, image input, and the Agent node's full tool loop.

Settings ▸ AI Providers. A ChatGPT (Subscription) card needs no key — only the CLI, installed and signed in.
Settings ▸ AI Providers. A ChatGPT (Subscription) card needs no key — only the CLI, installed and signed in.

Set it up

  1. Run npm install -g @openai/codex in Terminal.
  2. Run codex login and sign in with your ChatGPT account.
  3. Add the provider: Settings ▸ AI Providers ▸ Add Provider ▸ ChatGPT (Subscription).

The editor shows no key or URL field. In their place sit three live checks — Codex CLI installed (green shows the version; red offers the copyable install command), Signed in (red offers codex login), and Ready. Recheck re-runs all three at once.

Models

Seven models are available. They share one codex exec pipeline, so structured output and vision work on all of them. New nodes default to gpt-5.5.

ModelReach for it when
gpt-5.5The default — reasoning, nuanced generation, Agent runs
gpt-6-astraThe strongest reasoning, coding, and long-context work. Needs Codex CLI 0.153.1 or newer; an older CLI refuses it and says so
gpt-5.6-solThe most capable of the gpt-5.6 generation
gpt-5.6-terraEveryday prompts, balanced for speed
gpt-5.6-lunaClassification, extraction, short structured tasks
gpt-5.4A faster, balanced older option
gpt-5.4-miniThe lightest and fastest

The -pro variants are absent on purpose: the Codex CLI refuses them on a ChatGPT account.

Good to know

This is not a local provider. The codex binary is on your Mac, but it ships every prompt to OpenAI under your account. For nothing-leaves-the-machine flows, use Ollama, LM Studio, or Apple Intelligence.

The limit is shared. Calls draw on the same plan usage as the IDE extension, the CLI, and cloud tasks. A high-frequency trigger firing AI nodes can eat the budget you rely on elsewhere — put the busy, low-stakes steps on a local provider.

Watchflows adds no charge. AI cost tracking reports this provider as free, because there is no per-token bill to meter. Not affiliated with or endorsed by OpenAI.

Watchflows needs CLI version 0.142.5 or later, and gpt-6-astra needs 0.153.1. An older binary refuses Astra with a message that says so; the rest still run.

AI PromptAgentRun Script