rcli

A command line tool for running open models such as DeepSeek and Gemma on your own machine or hosted. You can chat, serve a model as an API, or open a coding agent on it.

Share on XLicense: MIT

Overview

Wally, from RunAnywhereAI, is a command line tool for running open models such as DeepSeek, Gemma and Qwen on your own machine, or hosted when a job outgrows it. One command lets you chat with a model, serve it as an OpenAI-compatible API, or open a coding agent on it. Local use needs no account or key.

Key features

  • Chat with a model using wally run
  • Serve an OpenAI-compatible API with wally serve
  • Pull models by name or any Hugging Face GGUF path
  • Open supported coding agents on a local model
  • Hosted models bill against RunAnywhere credit

Best for

People who want to try open models locally from a terminal. It supports Apple Silicon Macs, x86-64 Linux with glibc 2.35 or newer, and Windows, but not Intel Macs.

Upstream
RunanywhereAI/wally
Fork on GitHub
Guo-astro/rcli
Upstream stars
1.6k
Category
AI agents and LLM tools
Language
Rust
License
MIT
Forked
2026-03-08
Sync status
In syncLast synced 2026-09-29