Overview

Runars is a hosted inference proxy that brings Anthropic and OpenAI SDK compatibility to a catalog of high-performance models. The default catalog ships with GLM-5.2, a 744B-parameter Mixture-of-Experts model with a 131,072-token context window—operators can add more models over time, and GET /v1/models always reflects the current set.

What makes Runars different

• One catalog, two API shapes: Runars translates both Anthropic Messages API and OpenAI Chat Completions API requests to the underlying model you request. Your existing code works without changes—just point it at Runars and use a model id from the catalog.
• Canadian data residency: All inference runs on our Canadian GPU Cloud infrastructure in Montreal and Calgary. No data leaves Canada.
• Simple, predictable pricing: Pay only for what you use, per model. Credit-based billing with no contracts, and auto-reload keeps your balance from running out.

How it works

Your Application
  ├─ Anthropic SDK → POST /v1/messages
  ├─ Claude Code → POST /v1/messages  
  ├─ OpenAI SDK → POST /v1/chat/completions
  └─ Codex CLI → POST /v1/chat/completions
       ↓ (routed by requested model id against the catalog)
  Runars's translation layer
       ↓
  Requested model (Runars Canadian GPU Cloud, Montreal & Calgary)

Both the Anthropic and OpenAI wire formats are fully supported, including advanced features like tool calling, streaming, and prompt caching. Translation happens transparently—you use whichever SDK you prefer, and Runars handles the rest. Requesting a model id that isn't in the catalog returns 404 model_not_found rather than silently substituting another model.

Next steps

• New to Runars? Start with the Quickstart to make your first request.
• Setting up your tools? Jump to Install SDKs for Anthropic SDK, Claude Code, OpenAI SDK, and Codex CLI setup guides.
• Ready to dive into the API? Check out the API Endpoints reference.