Use MagmaRouter with SillyTavern
SillyTavern talks to any OpenAI-compatible API through its Custom source. Give it MagmaRouter's endpoint and a key, and the model list loads by itself.
- Base URL
- https://api.magmarouter.com/v1
- API key
- Get one, starts sk-mr-
- Example model
- glm-5.3-xploded
Connect
- In SillyTavern, open API Connections.
- Set the API to Chat Completion.
- Set the Chat Completion Source to Custom (OpenAI-compatible).
- Enter https://api.magmarouter.com/v1 as the endpoint. Use the API root, not /chat/completions.
- Paste your key. It starts sk-mr-.
- Pick a model from the fetched list, or type an id such as glm-5.3-flash-xploded.
- Click Test Message.
Which model
Every model on the line works here, and all but GLM 5.3 Xploded read images. glm-5.3-flash-xploded is the usual place to start. deepseek-v4.1-flash-1017-xploded and mimo-v2.6-flash-xploded are abliterated and rarely refuse; the uncensored models refuse far less than their base model and can still decline. The model list has every id.
What works
Model discovery through /v1/models, streaming, tool calling, vision on the models that read images, and reasoning all work through the same key.
