Axium-Labs/AX

A fast, lightweight CLI.

What it solves

Typical AI agents are heavy: they need a server, send telemetry, and take time to start. AX solves this by being a small, native terminal agent that stays on your machine. It starts fast, keeps sessions, memory, and config local, and requires no sign-up or remote service.

How it works

AX is built in Rust and ships as a single binary for Windows, Linux, and macOS. It loads everything on demand: skills, MCP servers, providers, and memory are only activated when needed, so startup stays minimal. You can start an interactive chat with ax, run a one-off task with ax run, or launch multiple independent tasks with ax agents. It discovers providers from your existing credentials and environment variables, and you can switch models anytime with /model. Your data lives in .ax next to the executable, with backup and import via ax export and ax import.

Who it's for

Developers and power users who want a fast, private AI assistant in the terminal without heavyweight infrastructure. It suits people comfortable with command-line workflows and who want to keep control of their data and model choices.

Highlights

  • Rust native: a single binary, no Node or Docker needed for core use
  • Fast startup: providers, skills, MCP servers, and memory load on demand
  • Local-first: sessions, memory, credentials, and config stored on your machine
  • Cross-platform: builds for Windows, Linux, and macOS from one source
  • Multiple providers: OpenAI Codex, DeepSeek, OpenAI, WorkBuddy, and OpenAI-compatible vendors
  • One-off and concurrent tasks: ax run for a single query, ax agents with bounded concurrency
  • Backup and import: export sessions and memory as .axpack files
  • Optional Computer Use with platform-specific adapters

Related

  • Project
  • Project
  • Project
  • Project