How to connect any LLM to SillyTavern
SillyTavern can talk to almost any model through one OpenAI-compatible endpoint. Here is the exact setup, how to switch models, and how to fix common errors.
SillyTavern is a front end, not a model host. To use it you point it at an OpenAI-compatible endpoint and paste a key, and then it can drive almost any model behind that endpoint. The cleanest way to reach many models is a gateway, so one key and base URL unlock the whole catalog instead of a single provider. Here is the exact setup, plus how to switch models and fix the errors people usually hit.
What you need first
Three things: a running SillyTavern install, an OpenAI-compatible endpoint, and a key for it. A gateway covers the last two at once, because it gives you one base URL and one key that reach many models. With UnoRouter you create an account, grab your key, and you are ready; there is a free tier so you can test a character before spending anything. Keep your key handy and do not paste it anywhere public.
The setup steps
In SillyTavern, open the API panel and choose Chat Completion, then the OpenAI-compatible (Custom) source. Set the base URL to https://api.unorouter.com/v1, paste your UnoRouter key, and click to load the model list. Pick a model from the dropdown and send a test message. Your characters, lorebooks, and presets work unchanged, because SillyTavern only needed an endpoint and a key. The same key also works in Janitor.AI, RisuAI, and Chub.
Switching models later
Because everything routes through one endpoint, changing models is just picking a different entry in the model dropdown; no new account, key, or base URL. That lets you run a cheap free model while you build a scene, then switch to a stronger paid model for the moment that matters. With 200+ models on one key you can match the model to the scene instead of settling for whatever a single provider offers.
Common errors and fixes
If the model list will not load, recheck the base URL ends in /v1 and the key is pasted with no stray spaces. A 401 means the key is wrong or expired; copy it again. A 404 on a model means that model name is not selected from the loaded list. If replies cut off, raise the response length in SillyTavern's settings. Most setup failures are a typo'd base URL or a key with whitespace, so check those first.
That is the whole trick
Connecting any LLM to SillyTavern comes down to one OpenAI-compatible endpoint plus a key, and a gateway turns that into many models behind a single setup. UnoRouter gives you that base URL, a free tier to start, and 200+ models you can switch between without redoing anything. Set it once and the same key follows you across every character chat front end.
Ready to connect SillyTavern? Create a free account or browse the models.
Nevika takes any OpenAI-compatible endpoint as a custom proxy. Here is the one minute setup, which model to pick, and the two errors people hit.
A roleplayer ranks three Claude Opus versions on UnoRouter: memory, emotional depth, drive, and creative writing. 4.8 wins, 4.6 beats 4.7, and each one wins or slips in a clear place.
SpicyChat is a zero-setup RP site whose memory and context are gated behind paid tiers. UnoRouter keeps the easy start but adds 200+ models you choose, deep lorebooks, and a key that also runs coding agents.