Skip to content
TrustedRails

Connect Open WebUI to TrustedRails

Open WebUI is a self-hosted, ChatGPT-style web interface for running models. It has built-in support for OpenAI-compatible endpoints, so you can add TrustedRails as a connection and chat with open-source models — no plugin required.

Prerequisites

Section titled “Prerequisites”

Add TrustedRails as an OpenAI connection

Section titled “Add TrustedRails as an OpenAI connection”

The first account you create in Open WebUI is the workspace admin. As that admin:

  1. Click your avatar (bottom-left) → Admin Panel, then open Settings → Connections.
  2. In the OpenAI API section, click Manage OpenAI API Connections, then the + (Add Connection).
  3. Fill in:
    • API Base URL — https://proxy.trustedrails.com/v1
    • API Key — your TrustedRails key (tr-prx-…)
    • Leave Prefix ID blank.
  4. Click Verify Connection — a success message means Open WebUI reached TrustedRails and fetched the model list.
  5. Save.

The models TrustedRails serves are now available to everyone in the workspace.

To connect just your own account instead of the whole workspace, add the same API Base URL and API Key under Settings → Connections → Direct Connections — no admin access needed.

Pick a model

Section titled “Pick a model”

Start a new chat and open the model selector at the top. TrustedRails's models — fetched from /v1/models — appear in the list. Pick moonshotai/Kimi-K2.6 (recommended) or MiniMaxAI/MiniMax-M2.7 (see Supported Models).

Verify

Section titled “Verify”

Send a message in the chat. A reply confirms Open WebUI is reaching TrustedRails through your key.

Troubleshooting

Section titled “Troubleshooting”
  • No models in the selector — re-open Admin Panel → Settings → Connections, click Verify Connection, and confirm the API Base URL is https://proxy.trustedrails.com/v1 and the connection is enabled.
  • 401 / invalid API key — wrong or paused key. Create a fresh one from Create a TrustedRails API Key.
  • Visible <think> reasoning in the output — reasoning models such as MiniMaxAI/MiniMax-M2.7 emit their thinking inline, which Open WebUI shows verbatim. Use moonshotai/Kimi-K2.6 for clean output.