Choose your setup
Choose where you want to connect n8n.
More details
Sign in on this computer first. Review the plan before any write.
More details
Grok signs in on your server. ChatGPT sign-in is not used.
More details
Experimental. Not affiliated with or endorsed by Anthropic. Uses the official Claude Code that you sign in to yourself; Relmio never sees your Claude login. Personal use only; Anthropic's terms apply and this may stop working.
Sign in in your own terminal. Use Basic LLM Chain + OpenAI Chat Model, Use Responses API off, Max Retries 0, Timeout at least 135000 ms and Batch Size 1. Text only, one request at a time; not AI Agent tools or the n8n AI Assistant. Renew approval in the wizard within 7 days.
Need another hosting option? Compare hosting routes
Relmio opens ChatGPT sign-in on this computer and keeps each registration separate. It never imports your personal Codex login or asks for a Platform API key.
On a server, this transfers one selected registration to the user-controlled installation. The destination owns future refreshes; local access stops after transfer. Review the exact server and background use before approving.
Checking selected ChatGPT registration.
Using ChatGPT plan
Connection and consent details
Identity sign-in may succeed without permission to use your ChatGPT plan. Allow that permission in a separate ChatGPT consent step before using models. Email is a label, not a way to merge registrations.
Relmio requests the account's public model catalog. A listed model does not guarantee admission or a completed response. This route supports text requests using Responses; it does not support image generation, audio, video, Files uploads, or stored conversations.
OpenAI receives model requests. ChatGPT plan limits and credits apply to the selected account. This local setup does not grant hosted website chat access or switch to Platform billing.
Check your server
Check the server identity before you connect to the server.
Choose your n8n
Choose the n8n to connect. This step only looks.
More details about the server check
Relmio uses read-only Docker commands here. It does not run commands inside n8n, change its image, or restart its container.
Manage detected self-hosted n8n
Other companions: SuperGrok Claude Local model
Review a Relmio-managed companion. This does not edit, rebuild, stop, recreate, or restart n8n.
Manage the installed ChatGPT session
Read the owner on the verified server. A transferred registration cannot be signed out from this computer.
Choose the n8n container and network, then check the installed account.
Sign in again on this computer with the same ChatGPT account and workspace. ChatGPT treats it as a new Relmio connection, and the old connection's usage settings do not carry over. Then sign out the old sign-in here and review the replacement. n8n needs the new Relmio key afterward.
Plan and usage
Press Refresh usage to read this sidecar's request counts.
Image generation uses a separate Codex sign-in, the way Hermes Agent does it. It isn't the official Sign in with ChatGPT, and OpenAI doesn't document this route for other apps, so it can stop working at any time. OpenAI recommends an API key for automation.
- Images use your plan's Codex limits 3 to 5 times faster than text, then credits. Free plans can't use it.
- This server stores a full Codex sign-in for the account. Anyone with root or Docker access on it can read it.
- Device code sign-in must be on in ChatGPT security settings, or allowed by your workspace admin.
- Sign in with the same ChatGPT account as this sidecar. In n8n, choose the model gpt-image-2.
Enter this code on the Codex sign-in page:
Only enter this code at auth.openai.com. Never share it.
Models marked In n8n show up in n8n's OpenAI nodes. Relmio asks OpenAI's catalog as the newest Codex release, checked from npm about every 12 hours.
When on, the sidecar sends one short test request ('Reply with OK') to models in your catalog: now for up to 12 of them, then for each new model, and again once a day for a model that failed. A test that gets no answer is tried again after an hour. n8n lists a model only after it answers; until any model has answered, n8n shows the full catalog except models that recently failed. Each test uses a small amount of your plan.
n8n's AI Assistant can't list these models. To try one, choose Self-hosted or OpenAI-compatible endpoint in its settings, use the Base URL http://n8n-openai-oauth:10531/v1 with the same Relmio key as your n8n OpenAI credential, and paste a model ID. Only do this if nobody else uses this n8n: their Assistant chats would use your ChatGPT plan. Relmio hasn't tested the Assistant with this sidecar, and OpenAI hasn't said whether this use fits its Sign in with ChatGPT terms.
Review the plan
Nothing is written until you approve this plan.
- Server
- Docker network
- Not selected
- ChatGPT account
- Reviewed n8n / network IDs
- Private hostname
n8n-openai-oauth
The wizard will
- Create or update only
/docker/n8n-openai-oauth - Transfer one selected registration to its user-controlled runtime
- Build and start only the private sidecar
- Keep 31 days of request and token counts here
The wizard will not
- Edit or rebuild the n8n image
- Stop, restart, or recreate n8n
- Publish port
10531 - Create a Traefik route
Host key, build and request count details
SSH host key
Image generation (optional)
Uses a separate Codex sign-in, not the sidecar's Sign in with ChatGPT permission. OpenAI does not document this image route for other apps, so it may stop working. OpenAI recommends an API key for automation.
Leave this unchecked to install without image sign-in. If selected, Relmio starts Codex sign-in after installation. You still approve it on OpenAI's page.
- Images use your Codex limits about 3 to 5 times faster on average than similar turns without images, then credits. Free plans cannot use them. Codex usage (opens in a new tab)
- This server stores a full Codex sign-in, including a refresh token. Root users, Docker users and backups can access it.
- Anyone who can run workflows in this n8n can use your account for images. Turn this on only if nobody else uses it.
- Device code sign-in must be enabled in ChatGPT security settings or allowed by your workspace admin.
- Use the same ChatGPT account as this sidecar. Relmio does not verify that the accounts match.
- Image requests send your prompt and any reference images to OpenAI.
Relmio keeps the SSH connection open while you complete image sign-in.
Connect n8n
Copy these values into n8n on the same private network.
Assistant companion status
- Code Sandbox URL
-
- Code Sandbox API key Shown once
-
- SearXNG URL
-
n8n environment settings
Apply these values yourself. Relmio never edits its Compose file or restarts its container.
Using ChatGPT plan · Manage usage (opens in a new tab)
Manage usage (opens in a new tab)1. Add the Relmio client credential
- Base URL
-
http://n8n-openai-oauth:10531/v1 - One-time Relmio client key
-
- Organization ID
- Leave empty
- Custom header
- Off
This key is not an OpenAI OAuth token.
Plan and usage: Relmio counts the text requests n8n sends through it, but not model checks or images. See them later under Manage the installed ChatGPT session.
Image generation:
Enter the Codex sign-in code
Open the Codex sign-in page and enter this code. Only enter it at auth.openai.com. Never share it.
SSH is still connected for image sign-in. Keep this Relmio tab open until you finish or cancel.
2. OpenAI Chat Model for AI Agent or Basic LLM Chain
- Selected model
-
- Responses API
- On for Chat Model node version 1.3
Models listed for this registration:
Select a model from this account's catalog. A listed model can still be unavailable for a specific request or serving region.
HTTP Request node
- Method
POST- URL
-
http://n8n-openai-oauth:10531/v1/chat/completions - Authentication
- Generic Credential Type
- Generic Auth Type
- Bearer Auth
- Credential
Relmio sidecar- Bearer token
- Use the one-time Relmio client key shown above.
- Authorization header
-
- Content-Type
application/json- Send body
- On · JSON · Using JSON
- JSON body
-
n8n examples
The README has AI Agent, Basic LLM Chain, JSON, and cURL examples. Start with no built-in tools.
You can close this wizard with Control+C. The connection on your server keeps running.
n8n stays unchanged. No restart, rebuild or public port.
Back up your n8n workflows first. Setup still writes to your server.
Nothing is written until you approve the exact plan.