LarkquillAI

API

Connection details

Paste these into any app with an "OpenAI-compatible" or "Generic OpenAI" option, like AnythingLLM, LibreChat, Open WebUI, or the OpenAI SDKs.

Base URL
API keyYour wby_... key from the API keys section, sent as Authorization: Bearer wby_...
Chat endpoint
Models endpoint
StreamingSupported ("stream": true, server-sent events)
Model nameContext windowDaily tokensPlan
Loading models…

Quickstart

Copy, set LARKQUILL_API_KEY, and run.

The API is OpenAI-compatible. Base URL . Models: quill-wren (Quill Wren 1.2, fast and lightweight), quill-scribe (Quill Scribe 1.1, everyday default), quill-sage (Quill Sage 1.5, advanced reasoning), quill-sovereign (Quill Sovereign 0.5, most capable flagship), quill-monarch (Quill Monarch 0.3, frontier reasoning). Every model is free; daily token limits apply. Older IDs (larkquill-chirp-1.1, larkquill-trill-1, larkquill-aria-1.5, larkquill-chorus-0.7, wrenby-*) and the old base URL still work.


  

Limits per key: 20 requests a minute and 300 a day (Quill Sovereign 0.5 counts as 3, max 30 a day). Errors use the OpenAI format: 401 bad key, 400 bad request, 429 rate or token limit.