Getting started

This is the 5-minute path from “nothing installed” to “a workflow ran on my box.” If anything breaks, see Troubleshooting.

Prerequisites

• macOS or Linux (x86_64 or aarch64).
• Ollama for local LLM use (recommended for first contact). Cloud providers — OpenAI, Anthropic, Mistral, Gemini — work too; set the right key in .env.
• (Optional) podman if you want sandboxed directory sessions.

1. Install Axocoatl.

   Curl (recommended)

   curl -fsSL https://axocoatl.ai/install.sh | sh

   Cargo

   cargo install axocoatl-cli

   From source

   git clone https://github.com/axocoatl/axocoatl && cd axocoatl
   cargo build --release

2. Pull a model.

   ollama serve &
   ollama pull llama3.2

3. Run axocoatl doctor to confirm everything is reachable.

   axocoatl doctor

   You should see ✓ on Rust toolchain, config, Ollama, model, data dir, and (if installed) podman.

4. Start the daemon.

   axocoatl dev

   This brings up the daemon, IPC, and the dashboard on http://localhost:8080.

5. Open the dashboard and fire a workflow.

   In your browser, open http://localhost:8080. Switch to Studio. In the left rail, expand Quick fire and click ▶ Fire all workflows. Watch the lattice cascade through every agent on the canvas.

What just happened?

The daemon spawned a supervised actor for every agent in your config. The lattice published a WorkflowStarted event. Every agent that lists the workflow’s entry point in its depends_on chain woke up and ran. Their outputs published TaskCompleted events; downstream agents activated as their thresholds were crossed. Nobody scheduled anything.

Open a directory session to see the agent actually read, write, and execute commands inside a sandboxed copy of one of your folders.

Next

• The event lattice — the coordination model.
• Agents — actor lifecycle, supervision, budgets.
• Skills — how the lattice routes work automatically.
• Sessions — Claude-Code-style directory work, locally.