SillyTavern API settings: How to connect as OpenAI-compatible
We explain the connection endpoint, key, model name, and context values in the SillyTavern API settings, along with verification commands.
Updated
Key points
- Select Chat Completion custom (OpenAI-compatible) for the connection method
- URL is https://api.mukenetsuapi.com/v1, model name is uncensored
- Context can go up to 100,000, but it is realistic to limit it for cost and speed
- Suspect the key for 401, the URL for 404, and parallel requests or rate limit for 429

Prepare 3 prerequisites first
SillyTavern API settings are short, but most stumbling blocks are due to incorrect input values. Please have the following 3 items ready first.
- API key: Issue it on the API key acquisition page. After signing up, you can receive $0.50 of trial credit with one click.
- Base URL:
https://api.mukenetsuapi.com/v1 - Model name:
uncensored
SillyTavern is a frontend that allows adding OpenAI-compatible endpoints. Since this API is OpenAI-compatible, you do not need to select the official OpenAI; you can register it as a custom endpoint.
Connection steps
Menu labels may vary by version. We describe the general flow here. If yours differs, please also check the SillyTavern documentation.
- Open the "API Connections" (plug icon) at the top.
- Select "Chat Completion" as the API type.
- From the source list, select the OpenAI-compatible custom connection (Custom).
- Enter
https://api.mukenetsuapi.com/v1in the endpoint URL. Do not append/chat/completionsto the end. - Paste the issued key into the API key field. Ensure there are no leading or trailing spaces.
- Set the model name to
uncensored. Select it from the list if available, or type it manually. - Press the connect button and verify that a success message appears.
Before connecting, verify the key and URL in the terminal to make troubleshooting easier.
curl https://api.mukenetsuapi.com/v1/chat/completions \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model": "uncensored", "messages": [{"role": "user", "content": "こんにちは"}], "max_tokens": 50}'If a response is returned, the key, URL, and model name are correct. If SillyTavern fails in this state, suspect the app-side input.
Set context and response length
The context for this API can use up to 100,000 tokens combined for prompt and output. However, we do not recommend setting SillyTavern's context size to the maximum. Each time, you resend the character settings, world view, and history, so larger values increase input per request.
Cost calculation example
If you use up to 16,000 tokens for context and the response is 300 tokens, the cost per message is as follows.
- Input: 16,000 ÷ 1,000,000 × $0.15 = $0.0024
- Output: 300 ÷ 1,000,000 × $0.50 = $0.00015
- Total: $0.00255
It costs $0.255 for 100 messages. Halving the context to 8,000 roughly halves the input cost. See the pricing page for unit price details.
Guidelines
For long stories or multi-character scenes, it is realistic to start around 8,000 to 16,000 and increase if the character seems to forget past events. Set the maximum response length to match the length you want to read as a single reply. max_tokens is optional, but setting a limit in SillyTavern is reassuring if you want to avoid overly long responses.
Sampling settings approach
temperature, top_p, stop, seed, and various penalties are passed directly to the API. You can adjust them using SillyTavern presets.
- Repetitive phrasing: slightly increase frequency or presence penalties.
- Rambling response: lower the temperature.
- Monotonous: slightly increase the temperature.
- Comparison: Fix the seed and change values one by one.
Preset item names vary by SillyTavern version. If settings do not take effect, check if your connection settings support the item or refer to the SillyTavern documentation.
Common errors and fixes
| Symptoms | Possible causes | Fixes |
|---|---|---|
| 401 | Wrong key or extra spaces | Re-paste the key and test it with curl. |
| 404 | Incorrect URL | Check that it ends with /v1. |
| Model not found | Model name spelling | Enter it in lowercase as uncensored. |
429 (concurrency) | Too many concurrent requests | Reduce rapid swiping or simultaneous generation in group chats. |
429 (rate_limited) | Exceeded requests per minute | Wait a moment. |
429 (with Retry-After: 5 header) | Temporary congestion | Wait 5 seconds and retry. |
Full keys have a limit of 8 concurrent, 300 per minute; trial keys have a limit of 2 concurrent, 60 per minute. Streaming responses use 1 concurrent slot until completion. Running group chats with trial keys will quickly fill the slots.
You can check if your balance is sufficient with the following command.
curl https://api.mukenetsuapi.com/v1/balance \
-H "Authorization: Bearer YOUR_API_KEY"You can also read it from the response header X-Balance-USD. To check the model list, use GET /v1/models.
curl https://api.mukenetsuapi.com/v1/models \
-H "Authorization: Bearer YOUR_API_KEY"Usage notes
This API is text-only. If you want to use SillyTavern's image generation or voice features, you need another service. The content is intended for fiction such as novels, roleplay, and companion apps within the scope of legal adult use. It will not generate sexual content involving minors in any setting. The prompt is not used for model training.
For more detailed parameters, see the documentation. Payments are made in USDT (TRC20) or USDC (Base) in integer amounts from $20 to $500.
Other guides
Frequently asked questions
What model name should I use in SillyTavern?
uncensored. Enter it in lowercase as is.
Do I need /chat/completions at the end of the URL?
Not needed. The base URL is /v1, and the app adds the remaining path.
Is a trial key sufficient?
It is sufficient for connection checks and short conversations. However, trial keys are limited to 2 concurrent and 60 per minute, and are valid for 7 days, so charge a regular key if you want to use group chats.
Can I also generate images?
No. This API is text-only and does not support image, voice, or video generation.
What should I set the context to?
The maximum is 100,000 tokens, but starting with 8,000 to 16,000 is practical. Larger values increase input per request, raising costs.
Get your API key with a single form
Create an account, copy your key, and change the base URL. That is all you need to complete the setup.