antrix.dev
Free guide

Setting Up GitHub

Create your GitHub account, set up repositories, and learn the basics of version control for your projects.

Reading this alone?

Do it with peers — bi-weekly calls, $29/mo founding.

Join →

What it does

GitHub hosts your code, tracks every version of every file, and manages your entire project. It's where your source code lives, where changes get reviewed before they go live, and where your app deploys from. It also doubles as your project management system — Issues track tasks and bugs, Projects organize them into boards, and Pull Requests are where code gets reviewed and merged.

Beyond version control, GitHub is a collaboration and automation hub. Pull requests give you structured code review — even when you're reviewing your own work, the diff view catches mistakes you'd miss in your editor. GitHub Actions handle CI/CD — run tests, lint code, and deploy automatically on every push. Issues turn your backlog into something structured and searchable. And the ecosystem of integrations is massive — Vercel deploys from your repo, Copilot suggests code in your editor, and third-party tools hook in through webhooks and the GitHub API.

Why we use it

GitHub is the industry standard and everything integrates with it. Vercel auto-deploys when you push code. Supabase can link to your repo. Claude Code works directly with git to commit and branch. Your code is backed up in the cloud, versioned so you can roll back any mistake, and accessible from anywhere. Even as a solo builder, git discipline saves you from yourself — every broken change is one git revert away from being undone.

The network effect seals it — every developer tool assumes GitHub. Your Vercel project deploys from it. Claude Code commits to it. Stripe webhooks can be tested against branches. Your CI pipeline runs on GitHub Actions. And your GitHub profile IS your developer portfolio — contribution history, pinned repos, and open source work tell a story that no resume can match. Choosing GitHub isn't just picking a host for your code — it's plugging into the infrastructure that every other tool in your stack already expects.

Setup checklist

1

Create a GitHub account

Go to github.com and sign up. The free tier is all you need — it includes unlimited public and private repos, GitHub Issues, Projects, and Actions. Use your professional email. Choose a username you'd put on a resume — it's your developer identity and shows up in every project URL, commit, and contribution.

2

Install Git locally

On macOS, the easiest way is through Homebrew: brew install git (install Homebrew first from https://brew.sh if you don't have it). Verify with git --version. Then set your identity so commits are attributed to you: git config --global user.name "Your Name" and git config --global user.email "your@email.com". Use the same email you signed up to GitHub with — this is how GitHub links commits to your profile.

3

Set up SSH keys for GitHub

SSH keys let you push and pull code without entering your password every time. Run ssh-keygen -t ed25519 -C "your@email.com". Press Enter to accept the default file location and set a passphrase if you want extra security. Copy your public key: cat ~/.ssh/id_ed25519.pub. Then go to GitHub → Settings SSH and GPG keys New SSH key, paste it in, and save. Test the connection: ssh -T git@github.com — you should see a success message.

4

Install GitHub CLI

Run brew install gh, then gh auth login. Choose SSH as your protocol and authenticate through your browser. The GitHub CLI lets you create repos, open pull requests, manage issues, and check deploy status — all from your terminal. It's faster than clicking through the web UI and integrates cleanly with your development workflow.

5

Create your first repository

From the terminal: gh repo create my-project --public --clone. This creates a repo on GitHub and clones it locally in one step. Or create it on github.com with a README, then clone: git clone git@github.com:username/repo.git. Either way, you end up with a local folder connected to a remote GitHub repo. The remote is your backup and your deployment source.

6

Learn the git workflow

The daily cycle is four commands. git pull — always pull before starting work to get the latest changes. git add . — stage your changes. git commit -m "what you changed" — save a snapshot with a message explaining why. git push — upload to GitHub. Check your status anytime with git status to see what's changed, what's staged, and what's not tracked. Run git log --oneline to see your commit history.

7

Set up branch protection

Go to your repo on GitHub → Settings Branches Add rule for main. Check Require a pull request before merging. This prevents you from accidentally pushing broken code directly to production. Every change goes through a pull request, which gives you a chance to review the diff, run automated checks, and catch mistakes before they reach your live app.

8

Use GitHub Issues for task tracking

Create an issue for every feature, bug, and task. Go to the Issues tab and click New issue. Add labels like bug, enhancement, priority: high to categorize them. Reference issues in your commit messages: “fix: resolve checkout error #12”. Close them automatically from pull requests by writing “Closes #12” in the PR description. This creates a clean trail from task to code change to deployment.

9

Set up a .gitignore

Every project needs a .gitignore file that tells git which files to exclude. For a Next.js project, you need at minimum: node_modules/, .next/, .env, .env.local, and .env*.local. Never commit secrets (API keys, database URLs), dependencies (node_modules), or build output (.next). Set this up before your first commit — retroactively removing files from git history is painful. Use gitignore.io to generate templates for your stack.

10

Learn to use Pull Requests

Pull requests (PRs) are how you merge changes safely. The flow: create a branch git checkout -b feat/my-feature, make your changes, commit them, push the branch git push -u origin feat/my-feature, then create a PR with gh pr create. The PR shows the full diff of your changes, lets you add a description of what and why, and creates a space for review. Once you're satisfied, merge it into main. This is the professional workflow used by every engineering team.

Pro tips

Explain WHY in commit messages — “fix: prevent double charge on retry” not “fix bug.” The diff shows what changed; the message explains why.

Use conventional commits feat:, fix:, chore:, refactor:, docs: prefixes keep your history readable and enable automated changelogs.

Create a PR template — add one at .github/pull_request_template.md to standardize what every PR describes: summary, test plan, and screenshots.

Star repos you reference often — they show up in your quick access and build a public trail of what you're learning.

Starter .gitignore for Next.js + Supabase

# Dependencies
node_modules/
.pnp.*

# Next.js
.next/
out/

# Environment
.env
.env.local
.env.*.local

# Supabase
supabase/.temp/

# IDE
.vscode/
.idea/

# OS
.DS_Store
Thumbs.db

# Debug
npm-debug.log*

# Vercel
.vercel

AI prompt to get started

I'm setting up a new GitHub repository for [project name]. Help me create: 1) A comprehensive .gitignore for a Next.js + Supabase project, 2) A PR template, 3) Issue templates for bugs and features, 4) Labels I should create for project management.

Mistakes to avoid

  • Committing .env files or API keys — use .gitignore and rotate any keys that were ever committed, even briefly
  • Giant commits with 20 files changed — commit small, commit often, one logical change per commit
  • Working directly on main — always use feature branches, even for small changes, so you can review before merging
  • Not pulling before starting work — leads to merge conflicts that are confusing and time-consuming to resolve
  • Ignoring .gitignore — set it up before your first commit, not after you've already pushed node_modules
  • Not using GitHub Issues for project management — writing todo lists in notes apps instead of Issues means your AI tools can't see or work on your backlog

Reading this and want to actually do it with others?

The community is bi-weekly live calls, async accountability, and a group of builders shipping in parallel. Founding seats are $29/mo.

[ 01 ] the community · waitlist open

Build next to people who actually ship.

AI isn't going to make you money. Products do. Clients do. Real businesses do. AI is the unfair advantage — wire it into how you build, sell, and serve, and you move faster than everyone still doing it by hand.

operator:youwaitlist:openfleet:yours to installcalls:weekly liveprice:$29/mo founding

Is this you?

$ qualify --you

  • You live in Claude Code (or Codex) all day — and you're still the one doing every task, one prompt at a time.
  • You bill by the hour, and you can feel the ceiling. You want agents doing the repeatable work.
  • You're shipping your own thing — and want the whole business automated, not just the code.
  • You're senior and employed — and want this leverage running before your industry expects it.
  • You'd rather install the system than watch one more video about it.
  • Not for the make-money-online crowd. Not for day-one beginners. This is for people who already ship.

What you get

os/

Antrix OS — your fleet's control room

The system I run on real client work, installed where you build — not a video walkthrough.

Claude Code core MCP servers + hooks subagents + worktrees one dashboard

skills/

ax-* skill packs + installers

Drop-in capability. The packs and one-command installers that do the repeatable work.

  • ax-audit, ax-deliver, ax-simplify, ax-content
  • Ship Your Product — installer
  • Run the Machine — installer
  • Overnight pipelines: issues → PRs

setups/

Wired-up service setups

The integrations already configured the way production needs them.

  • Stripe — billing + webhooks
  • Supabase — auth + data
  • Vercel — deploy + edge
  • Resend — transactional email

access/

Direct access — to me and the network

When you're mid-build and stuck, you post and get an answer from someone who's shipped it.

  • Live working calls every week — builds, teardowns, Q&A
  • Architecture + payment reviews on your code
  • A peer network of operators who ship, not spectate
  • First look at overflow client work
Antrix OS — projects across every company, with live agents and open tasksAntrix OS — the agent roster: system agents, PMs, personas, and workers

← drag to explore the dashboard

The control room you'll run your fleet from — every project, agent, skill, schedule, and secret in one place. Members get the codebase.

$ fleet --ships

IG carousels, drafted overnightLead lists scraped for consultingClient reports + invoices, generatedContent research, done by morningOvernight PRs from open issuesWeekly ops digest, auto-assembled

[ 03 ] work with me directly

Hire me, and you get me — not an agency.

Don't want to run the fleet yourself — or need senior hands on your product now? I take a few clients each quarter, working direct on your codebase, your stack, your business. No junior handoffs, no account managers. Best fit if you're already shipping something real.

You're stuck on the hard part

We pair and ship it — the product, the agent, or the integration that's been blocking you — in the session, not in a follow-up doc.

AI is still just a chat window for you

I wire agentic workflows, multi-agent pipelines, and scheduled jobs into your business — turning AI from a tab you open into real leverage on revenue work.

You shipped it but you're not sure it holds

I review architecture, AI usage, and payments, then tell you what to keep, what to rip out, and what to ship next. Straight, no hedging.

You want to run it yourself

Sometimes you don't need me to run it — you need me in the room while you do. I set up the system and hand you the keys.

Specialty — payments & billing

If your product touches money, that's where I help most — subscriptions, marketplaces, and migrations at Babbel scale, Stripe-certified. Those engagements run through my consulting practice at invocation.io.

Tell me what you're building.

I reply within 24 hours.