Quick Start

Start by importing your existing skill files into axel's global directory:

You can also import an entire directory of skills:

Verify your imported skills:

Skills are stored in ~/.config/axel/skills, making them available across all your workspaces.

In an existing repository, initialize an axel workspace:

This creates a minimal AXEL.md in your project:

workspace: my-project

layouts:
  panes:
    - type: claude
      skills:
        - "*"

  grids:
    default:
      type: tmux
      claude:
        col: 0
        row: 0

Launch the Claude pane to see your skills in action:

You'll see output like this:

axel creates symlinks from your centralized skills to Claude's expected location. This happens for each supported AI (Claude, Codex, OpenCode, Antigravity) - same skills, different destinations.

When the session ends, axel cleans up the symlinks automatically. Your skills stay in one place, portable across all LLMs.

Update your manifest to include Claude on the left, and two custom panes stacked on the right:

workspace: my-project

layouts:
  panes:
    - type: claude
      skills:
        - "*"
    - type: shell
    - type: server
      command: npm run dev

  grids:
    default:
      type: tmux
      claude:
        col: 0
        row: 0
      shell:
        col: 1
        row: 0
        color: yellow
      server:
        col: 1
        row: 1
        color: blue

Start your full workspace:

axel creates a tmux session named after your project, with all panes arranged according to your grid:

When you're done, kill the workspace to clean up skills and terminate the tmux session:

axel cleans up symlinks and restores your environment:

For fullstack development, you can create elaborate layouts with multiple AIs and services:

workspace: fullstack

layouts:
  panes:
    - type: claude
      skills: ["*"]
    - type: codex
      skills: ["*"]
    - type: frontend
      command: pnpm dev
    - type: backend
      command: cargo watch -x run
    - type: logs_fe
      command: tail -f logs/frontend.log
    - type: logs_be
      command: tail -f logs/backend.log

  grids:
    default:
      type: tmux
      claude:
        col: 0
        row: 0
      frontend:
        col: 0
        row: 1
      logs_fe:
        col: 1
        row: 1
        color: gray
      codex:
        col: 2
        row: 0
      backend:
        col: 2
        row: 1
      logs_be:
        col: 3
        row: 1
        color: gray

This creates a workspace with two AIs working on different parts of your stack: