Use TRAE with llmrelay

Setup

1. 1 Go to Settings > Models to open the model management panel.
2. 2 Click the add button and choose the custom configuration option (not the preset provider option).
3. 3 Set the API format to the Anthropic /v1/messages specification (or the OpenAI /v1/chat/completions specification).
4. 4 Leave the full-URL switch off and enter the base URL https://api.llmrelay.dev, or switch it on and enter https://api.llmrelay.dev/v1/messages.
5. 5 Enter the model ID (e.g. claude-opus-5) and paste your API key.
6. 6 Click the confirm button; TRAE calls the provider API to validate the key before adding the model.

Config

# TRAE: Settings > Models > Add > Custom configuration
# API format:  Anthropic  (/v1/messages spec)   | OpenAI (/v1/chat/completions spec)
# Full URL off -> Base URL: https://api.llmrelay.dev        # OpenAI would be https://api.llmrelay.dev/v1
# Full URL on  -> Full URL: https://api.llmrelay.dev/v1/messages
# Model ID:    claude-opus-5
# API key:     ***

Smoke test

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

Click confirm in the add-model dialog; a successful connection adds the model to the list, and failures print the provider's error and logs. Then send one chat message on that model.

Gotchas

• !The full-URL toggle changes the meaning of the field: off means base URL, on means complete endpoint. Mixing them up is the usual 404 cause.
• !Model-series presets (GPT-5 / Gemini-3 / Deepseek-4) change prompts and hyperparameters; the Deepseek-4 preset disables multimodal input, so leave it on Default unless the model matches.
• !The image-input switch is on by default for custom models; turn it off if the relay model has no vision support, otherwise uploads will fail at request time.

Related integrations