GitHub is a free online service where you keep a backup of your marketplace code online, along with the complete history of changes made to it over time. Git is the corresponding program that records that history on your computer.
Version control systems like Git and GitHub are a secure store for your code, independent of your computer. They let you revert changes if something breaks your marketplace and share your code with others working on it, such as Sharetribe Experts.
Version control systems have been used for a long time by all kinds of software developers. This article is written for someone doing AI-assisted development without previous developer experience.
This article picks up at Step 3 of our tutorial on how to start AI-assisted coding on top of Sharetribe, when the Template already runs on your computer, but before you’ve actually customized it. It is written for the Claude desktop app with the Sharetribe Development Plugin. If you use another AI tool, share this article with it and ask it to adapt the steps. It describes the plugin as of November 2026. If your AI assistant's steps differ from this article, follow your AI assistant.
If you’re reading this with AI-assisted development in mind, please also check our articles about what you can and can't do with vibecoding and how to succeed with vibecoding on Sharetribe. You can also read our vibecoding vocabulary to get acquainted with related technical definitions.
Why your marketplace code needs version control
When your marketplace runs without custom code, Sharetribe keeps its code and history for you. Once you start customizing with code, you need to store your code and history yourself on a version control system like GitHub. Keeping them up to date is up to you.
On GitHub, your code and its history live in a repository, or repo. Your GitHub repository is:
A backup of your code that doesn't depend on your computer
The history of every change, so you can return to an earlier version
What Render, your hosting service, builds your staging environment and live marketplace from
Where a developer or Sharetribe Expert works when you hire one
Learn more about Git and GitHub terms from our vocabulary article.
Where your marketplace’s custom code lives
Your custom code exists in two places:
The working, local copy on your computer, which the AI assistant you use edits and you see in your web browser at “localhost:3000”
Your private repository on GitHub, which your AI assistant connects as “origin”
Your Console settings, users, listings, and transactions are in neither of these places. They live in Sharetribe, so a Console change never needs to be sent to GitHub, and losing a repository never loses marketplace data.
When your custom marketplace is live, for example hosted on Render, the service has built your marketplace based on your custom code on GitHub. And when you download/clone the Sharetribe Web Template code, the source is the Sharetribe Template repository. You never update Sharetribe’s template repository yourself, but it acts as the original source for your marketplace’s initial code.
What you do and what your AI assistant does
The plugin runs Git for you, so you don't need to learn to use the technical Git commands. However, you still have a few things up to set up and keep running:
Create a GitHub account and an empty private repository - more on the importance of this later in this guide.
Tell your AI assistant to "Connect your code to GitHub."
Sign in to GitHub in your browser when asked.
Approve the commands your AI assistant proposes, and read its commit messages, the short notes it writes for each saved change.
Make sure your AI assistant keeps committing and pushing your code changes to the repository regularly.
After you go live, decide when a change you've checked on your staging environment goes to your live marketplace.
The first three happen only once. Keep the permission mode at Ask permissions while you're learning, so that your AI assistant shows you each command before it runs. After you’re used to the workflow, and if you feel comfortable, you can switch it to automatic.
When a command includes reset, force, discard, delete, or overwrite, ask your AI assistant to explain what will be lost before you approve it. Learn more about the words to be careful around.
Connect your code to GitHub
Before you start, you need a GitHub account. If you don't have one yet, sign up on GitHub.
Create an empty private repository
On GitHub, create the repository that will hold your code:
Click the + sign in the upper-right corner of any page, then click on “New repository”.
Type a name for the repository, for example your marketplace name without spaces.
Choose “Private” as the visibility.
Leave “Add README” off, and don't add a “.gitignore” file or a license.
Click “Create repository”.
GitHub then shows a page of setup commands, which you can ignore in case you’re using AI-assisted development, unless you want to learn the technical commands yourself. Keep the repository empty; its contents will be filled by your coding assistant.
Sign in to GitHub when asked
The first push needs proof that the repository is yours, and how you give it depends on your operating system.
Important: Enter your GitHub password only on GitHub's sign-in page in your browser. If you're asked for a token, enter it only in the terminal. Never paste either one into the AI assistant chat.
On Windows
Git for Windows includes a sign-in helper, Git Credential Manager. On the first push, it opens a browser window where you sign in to GitHub. Windows then remembers the sign-in, and Git won't ask again.
If pushes later fail with an authentication error, for example after a password change, remove the GitHub entry in User Accounts > Credential Manager in the Windows Control Panel. Then ask your AI assistant to push again. Learn more about caching your GitHub credentials in Git.
On a Mac
On a Mac, Git needs a helper to sign in through the browser. GitHub recommends two: GitHub CLI, which is GitHub's command-line tool, and Git Credential Manager.
Let your AI assistant install the helper it suggests. If it asks you to run a sign-in command, use the integrated terminal of the Code tab.
A GitHub page then opens in your browser. If the terminal shows a one-time code, enter it on that page, and then tell Your AI assistant you're signed in.
Note: If your AI assistant asks you to create a personal access token, a long code that works like a password, ask it to walk you through the steps on GitHub. Treat the token like your Sharetribe client secret.
Tell your AI assistant to connect and push
In this tutorial, we assume you have already installed the Sharetribe Development Plugin from the earlier tutorial. In the “Code” tab on your AI assistant’s desktop app, open a session with your project folder that has the web template installed and type /start. Type "Connect your code to GitHub"
Your AI assistant should ask for your GitHub username and the repository name. Then it pushes, which means it sends your “main” branch, the primary version of your code, to GitHub. With “Ask permissions” mode on, you approve each command, and you sign in when the first push asks you to.
When your AI assistant reports success, open your repository page (that you created earlier) in your web browser and refresh it. You should see the Template's files and a Private label next to the repository name.
The file list includes .env-template, .env.development, and .env.test, which come with the Template and are fine to keep. A file named exactly .env should not be there; that one holds secrets. Having the .env file and its secrets in the repository does not cause technical problems so technically it is possible to include it as long as the repository is explicitly private; leaving .env out is a security practice, since everyone with access to this repository has access to the same secrets.
When you're ready to move your marketplace online to staging, continue with Step 3 of the initial setup tutorial by saying:
"Deploy to a staging environment on Render"
Your AI assistant writes a render.yaml file, which tells Render how to build your marketplace, and guides you through connecting Render to your repository.
Keep the repository private and secrets out
Three rules keep your code and credentials safe:
Keep the repository private, unless you have deliberately decided to publish your code to the public instead of just the direct collaborators.
Avoid letting the .env file or any other credential reach GitHub. If the repository is not private, "avoid letting" changes to "never let".
Never paste a GitHub token or password into the AI chat.
The .env file holds your Sharetribe client secret and other keys, and the Template's .gitignore file tells Git to leave it out. You should keep your client secret and all other keys stored in another secure location, such as a password manager.
Important: Treat anything that has ever been pushed to GitHub as public, because Git keeps every earlier version. If you accidentally push a secret to the repository, generate it again rather than only deleting it from the repository.
For the Sharetribe client secret, create a new application in Build > Advanced > Applications in your Dev environment, and delete the old one. Then ask your AI assistant how to put the new values in your .env file and in Render without pasting them into the chat.
Save and publish each change
Once GitHub and Render are connected, every change follows the same loop.
You ask for a change, your AI assistant edits the copy on your computer, and you check it at localhost:3000. It then commits and pushes the change, and Render deploys it to your staging environment. After you go live, there's one more step. You check the change on your staging environment, and then your AI assistant pushes it to your live marketplace.
Commit after every working change
A commit is a saved checkpoint in your project's history, with a short message describing it. With the plugin, your AI assistant commits and pushes as soon as you confirm that a change looks right.
To review the changed files before you confirm, click the change counter in your session, such as +12 -1. Learn more about committing after every working change.
What Render does after a push
Before you go live, Render watches the main branch of your GitHub repository and redeploys your staging environment every time a push lands there. Render calls this setting Auto-Deploy, with the default value On Commit.
Going live changes this. When you ask your AI assistant to go live, it creates a second branch called staging and helps you set up a second Render service for your live marketplace. From then on, your staging environment builds from staging, and your live marketplace builds from main. New changes go to staging first. Your AI assistant pushes them to your live marketplace only after you've checked them and said yes.
A rebuild takes a few minutes, and you can follow it on your service's Deploys page in the Render dashboard.
Note: Render builds from GitHub, not from your computer. A change that was never committed and pushed never reaches staging.
If a change works at localhost:3000 but breaks on staging, tell your AI assistant that it only happens on staging. The cause is often server-side rendering.
Roll back a broken change
AI-made mistakes often look right at first, so sooner or later you'll want an earlier version back. You can roll back your code through Git, and Render updates your online version to match it. Neither undoes Console settings, listings, or user data that changed in the meantime - you’re just reverting the code changes.
1. Find the last working version
On your repository page, click the commit count at the top right of the file list. Your newest commits are at the top, each with the message your AI assistant wrote, and the Template's history from Sharetribe follows below.
Click a message to see what that commit changed, with added lines in green and removed lines in red. Pick the last version that worked on your staging environment or at localhost:3000.
You can also ask your AI assistant:
"List the last five commits with their dates and describe in plain language what each one changed."
2. Ask your AI assistant to return to it
Then tell your AI assistant which version to return to, for example:
"Undo everything after the version called 'Add favorite button to listing cards' with a revert that keeps the history, and explain what will change before you do it."
A revert adds a new commit that undoes the broken changes, so nothing disappears from the history. Your AI assistant then pushes, and Render rebuilds staging with the restored version.
Important: If your AI assistant proposes a reset, a force push, or discarding changes instead, ask it what will be lost before you approve. Learn more about revert and roll back.
Note: You don’t need to do anything in Render. The revert is a new commit, so Render redeploys staging automatically. Render also has its own Rollback button, which restores an earlier deploy without a rebuild. However, it turns off automatic deploys until you switch them back on, so for staging, a revert is the simpler route. Learn more about rollbacks on Render.
How your AI assistant uses branches
A branch is a separate line of work in your repository. You don't need to create or switch branches yourself, because the plugin does it for you.
Before you go live, everything happens on the main branch. After you go live, your AI assistant does new work on the staging branch and moves it to main, your live marketplace, when you approve it. If you're unsure where a change is, ask:
"Which branch are you on, and has this change reached my live marketplace yet?"
To see your branches, open the branch menu above the file list on your repository page. It lists every branch that has been pushed, and selecting one shows its files without changing anything on your computer.
A pull request is a GitHub page for reviewing a branch's changes before they're merged. You only need one when another person reviews your work, such as a developer or a Sharetribe Expert.
Give a developer or an Expert access
A developer or Sharetribe Expert may need your code, for example for a review before launch or a Template update. To give them access on GitHub:
Open your repository and click the Settings tab.
In the Access section of the sidebar, click Collaborators.
Click Add people.
Type their GitHub username, select them, and confirm.
Once they accept the email invitation, they can push changes, and you keep ownership throughout.
Their pushes don't update the copy on your computer automatically; you need to pull them. Before you ask your AI assistant for your next change, say:
"Commit my current work, then pull the latest changes from GitHub and tell me what changed. Do not force anything."
Tip: To see their changes before they reach main, ask them to work on a branch and open a pull request. After you click Merge pull request on GitHub, pull the changes the same way.
To run the Template, they need their own credentials from Console, so invite them to your organization in Console instead of sharing your .env file. Remove their access on GitHub when the work is done. Learn more about how to hire a developer.
About Sharetribe's Template updates
Updating the Sharetribe Web Template means taking in the newest code changes that the Sharetribe team has developed. This allows you to, for example, take certain new Console featuresfunctions into use, when they are released. This is the one GitHub task we generally recommend you don't do yourself, especially if it has been a long time since you have downloaded or last updated the template from the Sharetribe repository; the changes may interact with your own custom code. To be clear, depending on the situation, it is possible to update the template by yourself, but it is a bit of a riskier operation than usual.
Why the Template on a custom coded marketplace does not update itself
No-code marketplaces receive every Sharetribe release automatically. Once your marketplace runs on your own copy of the Template, new releases have to be merged into your code, because they may conflict with your customizations.
That's why your AI assistant kept Sharetribe's repository connected as upstream - it allows future code changes to be incorporated to your code. New releases can be fetched from there and combined with your changes.
Sharetribe lists every release on the Template's releases page on GitHub. Updates are optional, so you can skip releases you don't need. Learn more about what Template updates cost in terms of work.
Why you should consider having an Expert do the update
A merge combines two versions of the code that have both changed. Where both changed the same part of a file, someone has to decide which change to keep. A wrong decision can silently remove a customization or, for example, break checkout, and an AI can propose resolutions that look right and aren't.
Important: Even though you can use any developer you wish, we recommend hiring a vetted Sharetribe Expert for Template updates.
You can still ask your AI assistant for a check that changes nothing:
"Check what changed in the latest Sharetribe Web Template release and tell me whether it affects my marketplace. Do not merge anything."
If the AI assistant informs you of changes you would like to take into use, and that they would not conflict with your marketplace customizations, it may be safe to update the template with the help of AI. It may also be possible to do other kinds of more simple template updates with AI, especially if you are using the best available models. However, this depends a lot on the level of customization of the marketplace and how long it has been since you last updated the template, since the changes add up over time.
If you hire a developer, they can follow the developer documentation on pulling in the latest upstream changes. When they're done, pull their changes to your computer as described above.
Wrapping up
You don't need to learn Git commands to customize code safely, but you do need to read what your AI assistant proposes before you approve it. Keep your secrets out of GitHub, commit after every working change, and generally speaking, consider leaving Template updates to an Expert.
Remember to check out our other articles on AI assisted coding, including how to start AI-assisted coding on top of Sharetribe for non-developers, what you can and can't do with vibecoding, how to succeed with vibecoding on Sharetribe, and our vibecoding vocabulary.
Need to hire a developer? Connect with a vetted Sharetribe Expert on the Experts Marketplace.
