Installing AEL

This page describes how AEL, the Agent Engineering Language, will be installed, so that you know what to expect. When the beta launches, this page will give the installation steps for each operating system.

Status

Planned for AEL Beta 0.0.1. AEL is not available yet: there is nothing to download or install until the beta launches.

When installers will be published

When the beta launches, installers will be published on ael.openeng.ai, official packages on pack.ael.openeng.ai and the full documentation here, on docs.ael.openeng.ai.

The launch date will be announced in the AEL updates feed. Follow it in any feed reader: it will announce the launch date, and then the launch.

What the installers will set up

Installers for Linux, macOS and Windows will set up the AEL toolchain: the ael command (command names may change before launch) and what it needs to build for the targets you choose. The toolchain will be self-contained: one install will take your code all the way to a native program.

The installers will also come with target management and a health check:

target management will add what a target needs, such as a microcontroller board (Setting up a target);
ael doctor will report the toolchain, the installed targets and anything that is missing.

Preview syntax — may change before launch

ael doctor

Planned targets lists the exact targets planned for Beta 0.0.1.

What else you will need

What else you need will depend on what you build:

Desktop and server programs will need nothing else installed. Programs will compile straight to native executables, with no interpreter or virtual machine to run and no second language to install.
Microcontroller boards will need their makers' flashing tools. You will flash and monitor supported boards from the command line with ael flash and ael monitor, which will use each board's standard flashing tools.
Your existing programs will need their own runtimes. When an agent calls your existing Python, JavaScript, Java or .NET programs as tools, they will run with runtimes you provide, locally or on a remote worker.
Models will be yours to choose. An agent that calls a model will need a model endpoint: a local model server, a major cloud provider, or an inference endpoint you run yourself.

When your program starts

Your lock file will pin the exact versions you built with, so builds repeat, even offline, and nothing will download when your program starts.

Until the beta launches

A first program shows what the first commands will look like.
The Overview describes what AEL will be.
The FAQ answers common questions about availability, platforms and packages.