Open WebUI + TOKNADO
How to connect Open WebUI to TOKNADO's OpenAI-compatible API
Open WebUI is built around standard protocols, and it connects to any server that implements the OpenAI-compatible API. That includes TOKNADO, so you can give your whole team a chat interface backed by one endpoint and one key.
The setup is a single connection in Admin Settings. Most problems come from the URL (a pasted /chat/completions) or from Docker networking, both covered below.
Start free trial Jump to setup
What you need
- A running Open WebUI and an admin account.
- Your TOKNADO endpoint and API key from the dashboard.
- Optionally a model ID, if the model list doesn't populate on its own.
Setup: Open WebUI with TOKNADO
Open the connections settings
In Open WebUI go to Admin Settings → Connections → OpenAI.
Add a new connection
Click Add Connection (the plus icon) and fill in the fields. Use the standard/compatible option, not Azure.
URL : https://toknado.up.railway.app/v1 Auth : Bearer API Key : YOUR_TOKNADO_API_KEY API type: Chat CompletionsHandle the model list
Leave Model IDs empty to auto-detect, or add specific IDs to restrict what users see. If your endpoint doesn't return a model list, add the model ID manually.
Model IDs: toknado/autoSave and chat
Save the connection, start a new chat, and pick the model from the selector. Send a short prompt to confirm.
If Open WebUI runs in Docker
localhostinside the container is the container itself. TOKNADO is a public endpoint so this rarely matters, but if you proxy through a local service, usehttp://host.docker.internal:<port>/v1or a hostname the container can reach.
Notes for real-world use
- Open WebUI can use models for background tasks (titles, tags, follow-ups) in addition to your visible chat, so one conversation can cost more than one request. Keep that in mind against the rolling hourly limit.
- Streaming and tool calling follow the OpenAI format; if a feature behaves oddly, test with a plain chat first.
- For a shared deployment, create the connection once as admin so users never see the key.
Troubleshooting Open WebUI + TOKNADO
| Symptom | What to check |
|---|---|
| Chat returns 404 | Remove /chat/completions from the URL. Use the API root only. |
| No models appear | The endpoint may not return a model list. Add the model ID manually in the connection's Model IDs field. |
| UI becomes unresponsive after saving | An unreachable URL can stall the settings page. Fix or remove the connection from the settings; see Open WebUI's docs on recovering from a bad connection. |
| 401 / invalid key | Re-copy the key with no trailing space and make sure Auth is set to Bearer. |
| Container can't connect | Use a hostname reachable from inside the container, not localhost. |
FAQ
Does Open WebUI work with any OpenAI-compatible API?
Yes. Open WebUI connects to any server implementing the OpenAI-compatible API through Admin Settings → Connections → OpenAI.
What URL do I enter?
The TOKNADO endpoint from your dashboard, ending at the API root (typically /v1), with no trailing slash and no /chat/completions.
Can I limit which models users see?
Yes. Add specific IDs to the connection's Model IDs filter instead of leaving it empty.
Can several users share one connection?
Yes, an admin-level connection serves all users. Mind the shared hourly limit on your TOKNADO plan.