Use Open WebUI with llmrelay

Setup

1. 1 Open Open WebUI in a browser and go to Admin Settings -> Connections -> OpenAI.
2. 2 Click + Add Connection.
3. 3 Set URL to https://api.llmrelay.dev/v1 and paste your relay key into API Key.
4. 4 If model auto-detection fails, add model IDs to the Model IDs (Filter) allowlist with the + icon.
5. 5 Click Save, then pick a relay model in the model selector.

Config

# Open WebUI: Admin Settings -> Connections -> OpenAI -> + Add Connection
# URL:     https://api.llmrelay.dev/v1
# API Key: ***
# Model IDs (Filter): claude-sonnet-5, claude-opus-5, gpt-5.6-terra   # only needed if /models auto-detect fails

Smoke test

Copy, paste, run. If you get JSON back with a message, you're good.

Save the connection and confirm verification passes, or (if it errors) confirm the allowlisted model IDs appear in the model selector, then send one chat message.

Gotchas

• !Connection verification calls the provider's /models endpoint with a Bearer token; a 400/401/403 there does not mean chat is broken, you just have to allowlist model IDs manually.
• !The URL must end in /v1. Tested 2026-07-28: https://api.llmrelay.dev/models (no /v1) answers 200 with our marketing HTML page instead of JSON, so verification fails in a confusing way rather than erroring cleanly.
• !Open WebUI deliberately implements protocols rather than vendor APIs, so anything non-standard needs a pipe or a middleware proxy.
• !Running Open WebUI in Docker against a host-local server requires host.docker.internal instead of localhost (not an issue for a public relay, but the same URL rules apply).

Related integrations