Installing the agent (very early)

Here’s the least intuitive secret of this whole journey: you install the code agent before everything else.

Most tutorials have you configure the machine for three hours, network, firewall, services, then install the AI at the end, like a reward. We do the opposite. You set up the agent now, while the system is still bare, and it’s the agent that will help you hook up Tailscale, Cloudflare, Ollama and the whole shebang. You go from “I’m typing commands I don’t understand” to “I describe what I want, I review, I approve.”

Claude Code, OpenCode or Codex: which one?

These are excellent command-line code agents. They read your project, write files, run commands, and loop until it works. This guide follows three of them at parity, with a tab for each whenever a step differs, and the difference between them fits in one line:

• Claude Code: Anthropic’s agent. You plug into the Claude models (top of the Quelle IA Code ranking, in French, as of the September 28, 2026 edition). Ultra-simple setup, top reliability on long tasks. Included in the Claude Pro and Max subscriptions, or paid per use through the API.
• Codex: OpenAI’s agent, backed by the GPT models (GPT-6 Astra leads the Agents ranking, in French, in the same edition). Included in ChatGPT subscriptions, from Plus to Pro, or paid per use with an API key. It can also run on a local model (codex --oss).
• OpenCode: the open-source agent. Same spirit, but you choose your engine: Claude, GPT, Grok, or a local model via Ollama. It takes API keys and several subscriptions (ChatGPT, SuperGrok, GitHub Copilot). Perfect if you want 100% local or to juggle providers.

The two other frontier vendors also have their own terminal agent, included in their subscriptions. If you already pay for one of them, these are first-rank options:

• Grok Build (xAI, grok command): ticked on every xAI plan, from the free one to SuperGrok Heavy, with limits that rise with the subscription. Install: curl -fsSL https://x.ai/cli/install.sh | bash, then grok login --device-auth on a machine with no browser.
• Antigravity CLI (Google, agy command): it replaces Gemini CLI, which since June 18, 2026 no longer serves free accounts or Google AI Pro and Ultra (Gemini CLI is still reachable with a paid API key and with a Gemini Code Assist license). Written in Go, it orchestrates several subagents in the background. Install: curl -fsSL https://antigravity.google/cli/install.sh | bash.

Prices, limits for each tier and usage rules for all four subscriptions: Frontier models on a subscription.

And if you want to explore more widely, the landscape is rich, and they all work on the same principle (read, write, run, loop):

• Aider: a much-loved open-source agent, built around git (every change = a clean commit), compatible with just about any model, local ones included.
• Pi: the minimalist alternative to OpenCode. Same open-source world, opposite philosophy: a bare-bones core you extend yourself. See the callout right below.
• Cline: an agent that lives inside VS Code rather than in the terminal, if you’d rather stay in your editor. (Its cousin Roo Code shut down in May 2026.)
• Cursor, Devin Desktop (formerly Windsurf): full code editors, built around AI, for anyone who wants an all-in-one experience.
• Orca: not one more agent, but an environment to run them in in parallel (an “ADE”, Agent Development Environment). See the callout right below.

One key, every model: unified APIs

When an open agent like OpenCode asks you “which provider?”, you could create an account with each vendor (Anthropic, OpenAI, Google…), juggle five keys, five bills, five quotas. Or plug in a unified API: a single service sitting in front of dozens of models, exposing one key, one bill, one format. You switch models without switching accounts, and you pay for everything in one place.

• OpenRouter: the best known. A single key gives access to hundreds of models (Claude, GPT, Gemini, Grok, Llama, Qwen, DeepSeek…), with automatic routing to the cheapest or fastest provider, fallback if one goes down, and a dashboard that tracks spend down to the token. Pay-as-you-go, a thin markup over wholesale price. Ideal for comparing ten models on the same task without ten sign-ups.
• OpenCode Zen: the OpenCode team’s own gateway: a set of models curated and benchmarked for agentic coding (85 options as of October 1, 2026, from frontier to small-and-fast, including a few free ones). Fewer choices than OpenRouter, but each one is vetted to behave well inside an agent (not just in a chat). Pay-as-you-go (wholesale price + processing fees), with configurable auto-reload and monthly cap. To connect: /connect in OpenCode → “OpenCode Zen” → paste the key.
• OpenCode Go: the same idea, but on a flat plan: $10/month ($40 for Go Plus) for low-cost open coding models. The plan has caps, counted in dollars of usage: each model gets a monthly allowance ($15 to $60 on Go), with at most 20% of it in any 5-hour window and 50% in a week (checked on October 1, 2026 on the OpenCode Go page). The smart pick if you code every day and prefer a fixed amount to a variable bill.

Node.js: no longer required, still useful

Good news: none of Claude Code, Codex or OpenCode needs Node.js. All three install through a script that drops in a standalone program. You can skip this section and go straight to the install.

Node is still a tool that many projects ask for: a JavaScript site, an npm package, Pi installed through npm. When that day comes, install it with nvm (version manager) rather than the system package, to avoid permission headaches and switch versions painlessly.

# Install nvm (Node version manager)
curl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.40.8/install.sh | bash
# Reload your shell (or close/reopen the terminal)
source ~/.bashrc
# Install the latest Node LTS
nvm install --lts
node -v   # v24.x as of October 1, 2026 (Node 26 is due to become LTS on October 28)

Install and launch

# Downloads and installs Claude Code into ~/.local/bin
curl -fsSL https://claude.ai/install.sh | bash
# Open a new terminal, then check
claude --version

claude

curl -fsSL https://opencode.ai/install | bash
# alternative, if Node is installed: npm install -g opencode-ai

opencode

# Installs Codex (running the same command again updates it)
curl -fsSL https://chatgpt.com/codex/install.sh | sh
# alternative, if Node is installed: npm install -g @openai/codex

codex

# shows a URL and a code to enter from your computer (beta)
codex login --device-auth

Your first real move: have it configure the machine

This is where the magic happens. Rather than chaining system commands yourself, you ask. Here’s a first concrete mission, to copy almost verbatim once into the agent:

Lay the foundations of memory right away

Before you leave, create the file the agent will read at every session: a CLAUDE.md (Claude Code) or AGENTS.md (Codex, OpenCode) at the root of your project. Two lines are enough to start, we dig into the topic in Memory files.

cd ~/mon-projet
cat > CLAUDE.md <<'EOF'
# Machine context
- Mini-PC Ubuntu 26.04, use: personal dev workshop.
- Always show sudo commands before running them.
- No secret in plain text in the code.
EOF