Download — Axon

Installation

Choose your platform

The runtime is a single binary. No complex installer, no system dependencies to manage. Download, extract, and run.

Windows

Linux

macOS

Download the binary

Download axon-v0.4.0-windows-x64.zip for Windows 10 / 11 (64-bit).

axon-v0.4.0-windows-x64.zip SHA256 →

Extract and add to PATH

Extract the zip into a folder of your choice (ej. C:\axon) and add that folder to the system PATH.

PS> [Environment]::SetEnvironmentVariable("Path", $env:Path + ";C:\axon", "User") copy

Verify the installation

Open a new terminal and confirm the runtime responds.

PS> axon --version copy

          Axon Runtime v0.4.0 · Cognitive Flow Oriented

          Compiled: C++17 · Platform: Windows x64

          Status: OK · 62/62 tests

Your first flow

Create a main.axon file and run it.

PS> axon run main.axon copy

Note: If you're using Ollama as your provider, make sure it's running on localhost:11434 before executing any agent. Download Ollama a ollama.ai.

Download using curl or wget

Download the compiled binary for Linux x86_64. Also available for ARM64.

$ curl -LO https://releases.axon-pl.org/v0.4.0/axon-v0.4.0-linux-x64.tar.gz copy

$ tar -xzf axon-v0.4.0-linux-x64.tar.gz copy

PATH installation

Move the binary to a location in your PATH, or use the quick install script.

$ sudo mv axon /usr/local/bin/ copy

$ chmod +x /usr/local/bin/axon copy

Quick install script

Alternatively, use the script that handles steps 1 and 2 automatically.

$ curl -fsSL https://releases.axon-pl.org/v0.4.0/install.sh | bash copy

Verify and install with Docker

If you prefer to run Axon in an isolated container.

$ axon --version copy

$ docker pull axon-pl/runtime:0.3.0 copy

Docker: The image includes Ollama pre-installed. Map port 11434 to use local models: docker run -p 11434:11434 axonlang/runtime:0.3.0

Download the universal binary

The universal binary runs natively on Apple Silicon (M1/M2/M3) and Intel without Rosetta.

% curl -LO https://releases.axon-pl.org/v0.4.0/axon-v0.4.0-macos-universal.tar.gz copy

Install with the quick script

% curl -fsSL https://releases.axon-pl.org/v0.4.0/install-osx.sh | bash copy

Allow the binary in Gatekeeper

On first run, macOS may block the binary because it isn't signed with Apple. Allow it with:

% xattr -dr com.apple.quarantine $(which axon) copy

Verify

% axon --version copy

Apple Silicon: The runtime takes advantage of all M-series chip cores natively. parallel {} distributes work between Performance and Efficiency cores automatically.

System Requirements

Windows

OSWindows 10 / 11 x64

Min RAM4 GB

Recommended RAM8 GB

GPU (optional)NVIDIA CUDA / AMD ROCm

Disk~50 MB (runtime only)

Runtime dep.None

Linux

OSUbuntu 20.04+ / Debian / Arch

Kernel5.4+

Min RAM4 GB

GPU (optional)NVIDIA CUDA 11.0+

Disco~45 MB (runtime only)

Runtime dep.None (static linked)

macOS

OSmacOS 12 Monterey+

ChipApple Silicon · Intel

Min RAM8 GB

Recommended RAM16 GB (M-series)

Disk~48 MB (runtime only)

Runtime dep.None

Note: If you're using agents with local models, you need Ollama installed. The models themselves can take anywhere from 2 GB (llama3.2:3b) and 40 GB (llama3:70b). The Axon runtime itself doesn't require a GPU — only if you want accelerated local inference.

Verification

SHA256 Checksums

Verify the integrity of the downloaded file before running it. Compare the file hash against the ones listed here.

axon-v0.4.0-windows-x64.zip a3f8e2c91b4d7e6f0c2a1b8d5e4f3c2a1b8d5e4f3c2a1b8d5e4f3c2a1b8d5e4

axon-v0.4.0-linux-x64.tar.gz b7c9d3e4f5a6b7c8d9e0f1a2b3c4d5e6f7a8b9c0d1e2f3a4b5c6d7e8f9a0b1c2

axon-v0.4.0-linux-arm64.tar.gz c8d0e1f2a3b4c5d6e7f8a9b0c1d2e3f4a5b6c7d8e9f0a1b2c3d4e5f6a7b8c9d0

axon-v0.4.0-macos-universal.tar.gz d9e1f2a3b4c5d6e7f8a9b0c1d2e3f4a5b6c7d8e9f0a1b2c3d4e5f6a7b8c9d0e1

# Linux / macOS sha256sum axon-v0.4.0-linux-x64.tar.gz copy

# Windows PS Get-FileHash axon-v0.4.0-windows-x64.zip -Algorithm SHA256 copy

Axon Runtime v0.4.0

Choose your platform

System Requirements

SHA256 Checksums

You have Axon. Start here.