Skip to main content
Wally hosts frontier-scale open models. You reach them with the wally CLI or with one API key against an OpenAI-compatible endpoint, so any OpenAI client works.

​
Start with the CLI

Install wally, then sign in. Sign-in opens the console in your browser, and the CLI stores its key for you.
Hosted commands carry --cloud. This starts opencode on a model we host:
Check what you have spent with wally account usage. Usage bills against prepaid credits you buy in the console.

​
Or call the API

Create a Cloud key in the console and point any OpenAI client at https://inference.runanywhere.ai/v1. Production Wally runs with zero data retention by default: it does not store prompt or completion bodies.

Quickstart

Install, sign in and send your first hosted request.

API reference

Models, chat completions, streaming and errors.

Integrations

Point Claude Code, opencode or another tool at a model.

Run a model locally

Smaller models on your own machine, without an account.

​
On-device SDKs

The RunAnywhere SDKs run models inside your own app, on the user’s hardware. They are open source and maintained on GitHub, with one C++ core behind every platform. Start with the overview or pick a platform.

Swift

iOS, macOS

Kotlin

Android

React Native

iOS, Android

Flutter

iOS, Android

Web

Browser

Electron

Desktop