Connect SillyTavern, one step at a time
Keep one action in view while you move between this page and SillyTavern.
Settings path confirmed. A recorded end-to-end SillyTavern test is still pending.
Base URL
https://microrouter.xyz/v1Step 1 of 5
Get your microrouter key
Open the key page in another tab. No account is required.
Key page
https://microrouter.xyz/keyStep 2 of 5
Open API Connections
Return to SillyTavern and select the plug icon.
SillyTavern path
Plug icon → API Connections
Step 3 of 5
Choose the Custom source
Set these two menus before you paste the endpoint.
API
Chat Completion Source
Step 4 of 5
Paste the endpoint and your key
SillyTavern adds the request path, so the endpoint must stop at /v1.
Custom Endpoint (Base URL)
https://microrouter.xyz/v1Do not add /chat/completions.
Custom API Key
Your mr-live-… keyGet keyStep 5 of 5
Connect and send a test
Press Connect. Choose the starter model from the list, then use Test Message.
Starter model
meta/llama-3.3-70b1 / 5
Optional: test the endpoint from a terminalUseful when SillyTavern will not connect
Set MICROROUTER_KEY in your shell first.
curl https://microrouter.xyz/v1/chat/completions \
-H "Authorization: Bearer $MICROROUTER_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"meta/llama-3.3-70b","messages":[{"role":"user","content":"ping"}]}'Choose a model
All models →Start with Llama 3.3 70B. The other rows are alternatives to try for roleplay.
meta/llama-3.3-70bGeneral-purpose prose candidate.deepseek/deepseek-v3Creative-writing candidate to evaluate.zai/glm-5Long-context conversation candidate.mistral/ministral-14bSmaller-model candidate for frequent retries.
Linked models are in the current live catalog. Paste IDs exactly as written, author/name.
If it does not connect
Connect succeeds but the model list is emptyclient-side
Check that the endpoint is exactly https://microrouter.xyz/v1. Remove any /chat/completions suffix.
404, HTML, or a parse error on every request400 · invalid_request
Almost always the base URL. Use https://microrouter.xyz/v1 — the bare host without /v1 also works, but never append /chat/completions to the base URL field; the client adds that path itself. Open the error reference →
401 on every request401 · invalid_api_key
Wrong or rotated key. Keys start with mr-live-; after a rotation the old key keeps working for 24 hours, then dies. Open the error reference →
Response stops mid-sentence with a credits message402 · insufficient_credits
Balance hit $0 mid-stream. The stream ends with a terminal error naming the exact charge for tokens delivered — never a silent close. Open the error reference →