# LLM API Router

> An OpenAI-compatible gateway in front of several LLM backends. Users hold their own
> router API keys; upstream provider credentials never leave the server. Every call is
> logged and optionally traced to LangSmith.

## For AI agents

- MCP server (recommended): https://llm.nyolab.ai/mcp — Streamable HTTP, OAuth 2.1 with dynamic client registration.
  Tools: list_models, chat, usage_summary.
- OpenAI-compatible REST base URL: https://llm.nyolab.ai/api/public/v1
  - GET  /api/public/v1/models
  - POST /api/public/v1/chat/completions   (streaming supported)
  - Auth: `Authorization: Bearer rk_live_...` — a router key minted at https://llm.nyolab.ai/dashboard
- OpenAPI spec: https://llm.nyolab.ai/openapi.json
- Plugin manifest: https://llm.nyolab.ai/.well-known/ai-plugin.json

## Models

- qwen3-30b

## Pages

- https://llm.nyolab.ai/agents — agent setup and copy-paste client config
- https://llm.nyolab.ai/docs — full API documentation
- https://llm.nyolab.ai/playground — browser chat test bench
- https://llm.nyolab.ai/dashboard — mint and revoke API keys

## Access

Accounts require approval before API keys work. Sign up at https://llm.nyolab.ai/auth.