Roleplay APIOpenAI-compatible

One key.
Every roleplay app.

Paste one URL and one key into JanitorAI, Chub or SillyTavern. Roleplay fine-tunes and frontier models, streamed as they write.

Get an API key Read the docs

Works where you roleplay

JanitorAI and Chub take it as a reverse proxy. SillyTavern takes it as a custom OpenAI-compatible endpoint. Agnai, RisuAI and Open WebUI work too.

  • JanitorAI, Chub and SillyTavern, verified
  • Agnai, RisuAI, Open WebUI, LibreChat and more
  • Streaming replies over SSE

A key in a minute

Sign in, open the dashboard, name a key and copy it. Revoke it whenever you like.

  • Keys created and revoked in the dashboard
  • Sent as a standard Bearer token

Tune every model

Set temperature, top_p, top_k, min_p and repetition penalty per model, keep presets, and turn thinking mode on where a model has it.

  • Per-model sampler settings
  • A preset library
  • Usage stats in the dashboard

Every model has a raw twin

The plain name uses Kyrune’s tuned preset. Add -raw and your app’s own sampler values pass straight through.

  • Roleplay fine-tunes and frontier models
  • Pick either behaviour per request
Models
# GET https://kyrune.com/api/v1/models euryale-70b-v2.3        Kyrune preseteuryale-70b-v2.3-raw    your samplerhanami-70b-x1           Kyrune presethanami-70b-x1-raw       your samplerglm-5.3                 Kyrune presetglm-5.3-raw             your samplerdeepseek-v4-pro         Kyrune presetkimi-k2.6               Kyrune preset
Models
# GET https://kyrune.com/api/v1/models euryale-70b-v2.3        Kyrune preseteuryale-70b-v2.3-raw    your samplerhanami-70b-x1           Kyrune presethanami-70b-x1-raw       your samplerglm-5.3                 Kyrune presetglm-5.3-raw             your samplerdeepseek-v4-pro         Kyrune presetkimi-k2.6               Kyrune preset

Every model, one key

Roleplay fine-tunes on top, frontier models below. Each one also comes as a -raw twin.

Euryale 70B v2.3
Magnum v4 70B SE
Kunou 72B v1
Anubis 70B v1.1
Hanami 70B x1
Qwen 3.5 27B Anko
BlueStar v3 27B
DeepSeek V4 Pro
GLM-5.3
GLM-5.3 Flash Uncensored
Kimi K2.6
Qwen3.5 397B
MiniMax M3
MiMo V2.6 Pro
Mistral Large 3 675B
Nemotron 3 Ultra 550B
Hermes 4 70B
Step 3.7 Flash
Doubao Seed Character
LongCat 2.0

And much more

OpenAI-compatibleGET /models and POST /chat/completions.
Streams as it writesSSE until [DONE], and stopping a reply cancels it.
Thinking modeOn the models that support it.
PresetsSave sampler settings and switch between them.
Request a modelAsk for the fine-tune you want next.
Same accountImage Gen, K Agent and Files on the same login.

From key to first message

01

Create a key

In the dashboard after you sign in.

02

Paste it in

Base URL https://kyrune.com/api/v1 and your key.

03

Pick a model

Then say hello to your character.

Get an API key Read the docs

Common questions

https://kyrune.com/api/v1. JanitorAI also accepts the full /v1/chat/completions address.

JanitorAI and Chub as a reverse proxy, SillyTavern as a custom OpenAI-compatible endpoint, plus Agnai, RisuAI, Open WebUI, LibreChat, LobeChat, AnythingLLM, Chatbox and NextChat.

Every model has a -raw twin. The plain name uses Kyrune’s tuned preset and sampler; -raw passes your app’s own sampler values through.

Yes. A free account gets 10 requests a day. Paid plans raise the limits, and Sponsor and up have unlimited roleplay.

Up to three on kyrune.com.