Docs / Quickstart

Deploy a pack in one command

The autonomous packs are built to run on your own Debian/Ubuntu VM. The standard path installs system dependencies, creates a Python environment, detects an LLM CLI if available, and sets up a systemd service.

Prerequisites

• A Debian or Ubuntu VM with shell access.
• A pack from The Autonomous Stack extracted on the machine.
• Optional: an LLM CLI already installed and connected with your subscription.

Install

From the pack folder, run the bundled installer. The READMEs of packs such as SRE Agent, Release & Rollout Agent, and Free Sample Agent document this install form.

unzip agent-sre-v1.zip
cd agent-sre-v1
sudo bash install.sh

For a trial without a systemd service, several packs also expose a manual mode:

bash install.sh --no-service

Connect your LLM subscription

The model is BYO-LLM: the pack ships no LLM API key. When a compatible CLI is present and connected, install.sh or llm_adapter.py can detect it and the pack uses your existing subscription.

# Pick a single backend and connect it once.
npm i -g @anthropic-ai/claude-code
claude

npm i -g @openai/codex
codex login

# Optional: force the backend for compatible packs.
export PACK_LLM=claude   # or kimi, or codex

If no LLM CLI is detected, packs that advertise a deterministic core keep working without LLM reasoning. Enriched functions, such as diagnosis, adjudication, or written reports, are enabled only when a compatible CLI is connected.

Verify

Recent packs include offline tests and status commands. The exact names depend on the pack, but the following forms are documented in the existing READMEs:

bash smoke_test.sh
python agent.py status
python agent.py run --once

After install

Edit the pack's example files to describe your services, targets, or workflows. For example, SRE Agent uses targets.json, while Release & Rollout Agent uses services.json.