n8n-nodes-openai-codex

Community n8n AI Chat Model node: OpenAI codex.

Installation

Follow n8n community node installation docs:

• https://docs.n8n.io/integrations/community-nodes/installation/

Features

• Node name: OpenAI codex
• Type: AI Language Model root node (ai_languageModel)
• Device-code login flow compatible with Codex-style auth endpoints
• Auth state persisted on disk and reused across runs
• Automatic token refresh (proactive refresh before expiry)
• Dynamic model catalog from backend /models (cached on disk)
• Reasoning-effort options loaded from selected model capability
• Tool-calling support for n8n AI Agent (exact tool-name preservation)
• Explicit context strategy switch: Memory Only or Backend Chain (previous_response_id)
• Strict model-to-reasoning validation (invalid model/effort combinations are rejected)
• Runtime model-substitution guard (throws if backend returns a different model slug)

Login Flow

1. Add the OpenAI codex node.
2. Click Test step.
3. If login is required, the node output/error contains:
   • verification URL
   • user code
4. Complete login in the browser, then click Test step again.

Persistence

• Default auth state path: $N8N_USER_FOLDER/openai-codex-state
• Fallback path: ~/.n8n/openai-codex-state
• Override path: N8N_OPENAI_CODEX_STATE_DIR=/absolute/path

Context Strategy

• Memory Only (n8n Memory):
  • previous_response_id is disabled.
  • Use n8n Memory nodes (Simple/Redis/Postgres/Mongo/etc.) as context source.
• Backend Chain (previous_response_id):
  • Session Key is required.
  • The node stores and reuses previous_response_id per Session Key.
  • Use a stable key per conversation/user (for example ={{ $json.userId }}).

Use one strategy at a time to avoid redundant context.

Backend Defaults

• Base URL: https://chatgpt.com/backend-api/codex
• Originator header: codex_cli_rs

Model Identity Note

• Asking the model "what model are you?" is not a reliable identity check.
• Use backend metadata (response.model) instead.
• This node validates backend-returned model slug against the requested model and fails fast on mismatch.

Development

pnpm install
pnpm lint
pnpm build