New to GitHub? Your first AI pull request, start to finish.
You have never used GitHub. In one sitting you will create a project, ask an AI agent for a change in plain English, read what it did, ask for one adjustment, and approve the result.
Overview
Why this way is worth five extra minutes
You could paste a request into a chat window and hope. The way you are about to learn adds a few small habits, and they pay off on the first change you want to undo.
- You see the change before it counts. The agent's work is laid out line by line, and nothing changes in your project until you accept it.
- You ask for adjustments in plain English. A comment on the pull request brings back a new version.
- You can go back. GitHub keeps every version of your project, and a merged pull request has a Revert button that opens the undo as a new pull request.
Professional teams work this way for the same reasons. ProPR gives you the same loop with no setup on your computer: one website for your project, one for the agent, and your browser.
Step 1
Create a GitHub account and a repository
GitHub is the website where your project will live. Sign up at github.com/signup with an email address, or continue with a Google account. A free account is all this walkthrough needs.
Next, create a repository: a folder on GitHub that holds one project and remembers every change ever made to it. In the top-right corner of any GitHub page, press the + button and choose New repository. Give it a name such as my-first-site, tick Add a README file, and press Create repository. You can keep it private.
The README is a short text file that describes the project. It is the file you will ask the agent to change first, so there is something to see.
Step 2
Start a ProPR Hosted instance and connect it
ProPR is the service that turns your written requests into changes on GitHub. ProPR Hosted runs it for you, so there is nothing to install. Choose a Hosted plan: Solo is enough for one person. Sign in with GitHub when checkout asks, pick the GitHub account that owns your new repository, and pay.
Your instance opens in the browser. Sign in with the same GitHub account, open Repositories, and add the repository you created. From now on ProPR watches that repository for your requests.
The live demo is a full ProPR instance you can click through before you buy.
Step 3
Add an AI subscription
The agent that writes the code is Claude Code or Codex, and it runs on your own AI account. If you already pay for Claude or ChatGPT, you are set; otherwise sign up for one of them first. Providers bill you directly, and ProPR adds nothing on top.
In your instance, open Coding Agents, add Claude Code or Codex, choose Log in to a new account, and sign in with your Claude or ChatGPT account in the browser window that opens. The agent shows Ready when it is connected.
Step 4
Write an issue in plain English
An issue is a note on your repository that describes one thing you want done. On GitHub, open your repository, click the Issues tab, then New issue. Write a title and describe the change the way you would explain it to a colleague:
Before you submit, open Labels in the right-hand column and add the label AI. If the repository has no such label yet, type the name and GitHub offers to create it. Press Submit new issue. That label is the hand-off: ProPR picks the issue up within moments and the agent starts working.
Clear requests get good results. Say what the page should contain and who it is for. Leave the how to the agent.
Step 5
Watch the agent work
Back in ProPR, the task appears on your dashboard under Happening now, and GitHub adds an AI-processing label to your issue. Open the task to follow the live log: the agent explains, in plain language, what it is looking at, what it changed, and any checks it ran.
Behind the scenes the agent works on a branch: a separate copy of your project where changes are made without touching the main one. Each saved change on that copy is a commit. You do not need to manage either; ProPR does, and both become visible in the next step.
Step 6
Open the pull request and read what changed
When the agent finishes, ProPR opens a pull request: a request to bring the agent's changes into your project, with every change laid out for you to read. The issue's label changes to AI-done, and the task in ProPR links to the pull request. On GitHub, click the Pull requests tab and open it.
The Conversation tab holds a summary of what was done and why. The Files changed tab shows every line: green lines were added, red lines removed. For your first request you will see one new file, index.html. Nothing has changed in your project yet. The pull request is a proposal until you accept it.
Step 7
Ask for a change with an ordinary comment
Reading the change and asking for adjustments is called review, and you do it in plain English. At the bottom of the Conversation tab, write what you want different and press Comment:
Please make the heading larger and add a second sentence inviting people to come back soon.
ProPR reads the comment, starts a follow-up run, and adds one more commit to the same pull request. It replies with a link to the run, the new commit, and an undo link that removes only that follow-up if you preferred the first version. Repeat as often as you like, and each round adds a commit you can see and undo.
Step 8
Merge
When you are happy with it, press Merge pull request, then Confirm merge. To merge is to accept the proposal: the agent's changes become part of your project, and GitHub closes the issue for you because the pull request named it.
Open your repository's front page and index.html is there. Repositories that run automatic checks on every pull request, a step called CI, show a green tick above the merge button when the checks pass. Your first repository has none yet; the walkthroughs below add them.
From here on, the loop is the same: write an issue, read the pull request, comment, merge. Your project history keeps a record of each round.
Create a simple web page called index.html. It should say "Welcome to my first site" as a heading, with one sentence below it about what the site is for. Keep the styling minimal and friendly.