How to Use Replit: The Beginner's Guide (First Prompt, Agent 4, Plan Mode, and Your First Live App)

CodexFuse Team · 2026-10-28T17:59:35.206Z · 16 min read

Learning how to use Replit takes about ten minutes, and you do not need to write a single line of code to get there. Replit is a browser-based workspace where you describe the app you want in plain English and its AI, called Agent, builds it for you: it writes the code, sets up the database, tests the result, and publishes it to a live web address. This guide walks through the whole first-app flow the way Replit's own beginner guide teaches it, from signing up to a published link, plus the Plan mode habit and the Free mode budget that keep your first build free. If the idea of describing software instead of coding it sounds new, our guide to vibe coding explains the concept behind it first.

Replit used to be mainly an online code editor. In March 2026 it relaunched around Agent 4, an AI that turns a description into working software. That change is why the guides ranking today split into two kinds: older tutorials about writing code in an editor, and newer ones about prompting Agent to build an app. This guide follows the newer path, because it is the one Replit itself now teaches, and it notes where the classic editor still matters.

How do you use Replit?

To use Replit, create a free account, start a new chat, choose Website as the output, and select Free Mode. Describe one clear outcome, what visitors must be able to do, and what not to build yet. Replit's Agent builds the app, then you test it in Preview and ask for one improvement at a time.

Here is the whole flow in seven steps:

1. Sign up for a free account. Go to replit.com and create an account with an email address, Google, or GitHub. No installation is needed; everything runs in your browser.
2. Start a new chat and pick Website. Open a new chat, choose Website as the output, and open the mode selector to choose Free. If there is no output menu, use Add attachment, then Create something new, then Website.
3. Write one prompt with boundaries. Describe a single outcome, list the things visitors must be able to do, add success checks you can test, and name what not to build yet, such as logins, payments, or a database.
4. Let Agent plan first. Use Plan mode so Agent breaks the project into an ordered task list you can review. Approve the plan before any code is written.
5. Watch Agent build and test itself. Agent writes the code, sets up the infrastructure, and checks its own work. A first build in Free Mode can take a few minutes.
6. Test like a visitor. Click every page and button in Preview, submit the form, and resize the preview to a phone-sized width. Compare what you see against the success checks from your prompt.
7. Improve one thing at a time. Ask for a single clear change, keep the existing design, and test again after each change. When it works, click Publish for a live link.

What is Replit, in plain English?

Replit is an online platform for building software in your browser. You do not install anything, and your code is saved in the cloud, so you can open the same project from any device. It started as a code editor that supports many programming languages, and it is now best known for Agent, the AI that builds apps from a plain-language description.

Agent is not a chatbot that only answers questions. According to Replit's own documentation, Agent takes action: it sets up the project, creates applications, checks its work, and fixes problems along the way. You describe what you want in everyday language, and Agent handles the rest, from planning to deployment. No code or technical knowledge is required to start.

Agent 4, launched in March 2026, added three things that matter for beginners. First, an infinite Design Canvas where you can generate several design variants, compare them side by side, and apply your favorite to the running app. Second, parallel agents that split a complex request into smaller pieces and work on them at the same time. Third, the ability to build more than one thing inside a single project: a web app, a mobile app, slides, and a dashboard can share the same backend and data.

What can you build with Replit?

You can build web apps, mobile apps, data dashboards, and AI-powered tools. Replit's documentation lists the same range, plus visual designs and prototypes on the Design Canvas, and generated files such as CSVs, PDFs, PowerPoint files, and Markdown documents. Agent can also pull data from connected services such as BigQuery, Linear, Slack, and Notion directly from chat.

Good first projects are small and easy to test. A habit tracker, a class schedule page, a simple dashboard, a landing page for a side project, or a quote calculator all fit. Pick something you would actually use, because you will test it the way a real visitor would. Avoid starting with a full online store, a social network, or anything with payments and user accounts; those come later, after the core works.

Replit also has built-in services that need no setup: authentication with email and password or Google, a PostgreSQL database with a visual schema editor, object storage, hosting, and monitoring. You do not need to configure any of them for a first app. Agent sets up what your description asks for and leaves the rest out.

How to use Replit Agent to build your first app

Replit Agent builds your first app through a short loop: prompt, build, test, and refine. The quality of the result depends almost entirely on how you describe what you want, so most of this section is about the prompt. Follow the steps in order and resist the urge to add five features in one message.

Step 1: Create your account and open a new chat

Go to replit.com and sign up for a free account. You can use an email address, Google, or GitHub. Once you are in, start a new chat. This is the new way to build: you talk to Agent in chat instead of opening a blank editor. The classic editor is still there for writing code by hand, but your first app starts in chat.

Step 2: Choose Website as the output and Free Mode

Before you type anything, choose Website as the output so Agent knows what kind of thing to build. Then open the mode selector and choose Free. Free Mode uses intelligent model routing to pick the AI model for you, and Replit's beginner guide uses it to build a first app. Starting in Free Mode also means your first build costs nothing while you learn the loop.

Step 3: Write a prompt with boundaries

A good first prompt has four parts: one outcome, what visitors must be able to do, success checks you can test, and a list of things not to build yet. Here is the shape Replit's guide teaches, using a yoga studio site as the example:

Build a simple website for my local yoga studio where visitors can browse this week's classes, open a class to see the details, and request a spot with their name and email. Show a clear confirmation after they submit. Success: I can browse classes, open one, submit a request and see the confirmation. It works on a phone-sized screen. Do not add logins, payments, a database or any integrations yet.

The last line matters more than it looks. Postponing login, data, payments, and integrations keeps the first build focused, and you add them once the core works. A prompt without boundaries usually produces a bigger, buggier app that is harder to test.

Step 4: Use Plan mode before Agent writes code

Plan mode is the habit that prevents most "it built the wrong thing" frustration. Click Plan in the chat input, or simply ask Agent to enter Plan mode, and Agent will break your project into an ordered task list, explore different approaches, and weigh the trade-offs before any code is written. Review the list, ask questions, and request changes. When you are happy with the plan, approve it and Agent starts building.

Plan mode changes no code by itself. That is the point: you get to see the shape of the build and fix misunderstandings while they are still cheap. Start with Plan mode every time, even for small projects. It is the single most reliable habit in the community guides, and it is the step beginners skip most often.

Step 5: Let Agent build, then test like a visitor

Send the prompt and follow the progress in chat. Agent plans the app, creates it, starts it, and checks the result. Replit warns that Free Mode may take a few minutes for a complete first build, so give it time instead of sending a second message. When Preview is running, use the app the way a customer would: click through every page and button, submit the form and check you see the confirmation, and resize Preview to a phone-sized width.

If something is broken, describe the symptom to Agent, such as "the confirmation never appears after I click submit," rather than trying to diagnose the code yourself. That is Replit's own advice, and it works, because Agent can see the code and the error while you cannot. Every step creates an automatic checkpoint, a full snapshot of your code, database, and project state, so if a change makes things worse you can roll back to any previous state.

Step 6: Improve one thing at a time

Once the basics work, ask for a single clear change, keep the visual style, and say what not to add. For example: add a page showing the classes I have requested in this session, keep the existing design, and do not add login or a database. Test again after each change. Replit calls this the core loop: prompt, build, test, and refine.

If Agent recommends a higher mode for a more complex request, it is your choice whether to switch. Before you do, try splitting the request into two or three smaller steps and testing after each one, because smaller asks are easier to check and fix. Keep a short note of the prompts that worked; the ones that produced a good result are worth reusing on the next project.

What is Plan mode in Replit?

Plan mode is Replit Agent's planning step. Instead of building immediately, Agent breaks your project into an ordered task list, explores different approaches, weighs trade-offs, and lets you review and refine the plan before any code is written. You enable it by clicking Plan in the chat input or by asking Agent to enter Plan mode.

Plan mode exists because the most expensive mistake in AI building is building the wrong thing. A plan you can read and correct costs far less than an app you have to rebuild. The community guides put it bluntly: start with Plan mode every time, review the plan, ask questions, and only approve it when it matches what you meant. If the plan is wrong, say so in plain language and let Agent revise it.

One caution: on effort-based billing, Agent interactions can consume credits, and community guides warn that this includes planning conversations that produce no code. Watch your usage in Settings, and keep plans short and specific. Replit asks for confirmation before any paid action starts, and Agent returns to Free Mode afterward, so a paid step never happens silently.

Replit Agent modes: Free, Power, and Max

Replit Agent has three modes, plus an automatic option for paid work. Free Mode is no-cost everyday Agent work within your plan's allowance. Power is a paid, mid-tier mode with powerful models at an affordable price. Max is a paid, top-tier mode with the best frontier models for the most demanding work. Auto selects a model for each paid task in Power or Max.

Free Mode allowances reset every five hours on Core and Pro, and there are weekly limits. The Starter tier keeps its existing limits, which are smaller. If a request needs more capability than Free Mode, Agent may offer a paid Allow once option. Replit asks for confirmation before the paid action starts, and Agent returns to Free Mode afterward. Check your usage in Settings to see where you stand before a big build.

For most beginners, the rule is simple: stay in Free Mode until a specific request fails there, then split the request into smaller steps before paying for a higher mode. Smaller asks are easier to check and fix, and they usually succeed in Free Mode anyway.

How to publish your app on Replit

Publishing is the last step of the loop, and it is one click. When your app works in Preview, click Publish and Replit gives it a live URL immediately. There is no server to configure, no deployment pipeline to set up, and no domain to buy for a first version. Custom domains are supported on paid plans.

There is one free-plan catch to know before you share the link. On the free Starter plan you get one free published app, and Replit's documentation says your free published link automatically goes down after 30 days. Starter is fine for learning and testing ideas, but it is not the plan for a page your customers rely on. Move to Core before you send customers to anything, and remove the Made with Replit badge at the same time, since badge removal needs Core or above.

Before you publish, run the visitor test one more time on the live link, not just in Preview. Open it on your phone, click every button, and submit the form. The published app is the one people will actually use, so the last check should happen there.

Replit pricing: free vs Core vs Pro

Replit has three plans. Starter is free. Core costs $20 a month, or $18 a month when billed annually. Pro costs $100 a month, or $90 a month when billed annually. All prices are in US dollars, checked on Replit's pricing page on 28 October 2026. Prices can change, so confirm them on the pricing page before you pay.

The line between Starter and Core is the line between learning and shipping. Starter teaches you the loop and lets you publish one app to show people. Core removes the badge, keeps the app online, and gives you the Free Mode hours and project allowance to build seriously. Pro is for teams and heavy builders: 10 parallel agents, 10 active background tasks instead of Core's 1, and the collaboration limits a small team needs.

Do not buy a plan to learn. Build your first app on Starter, publish it, show it to someone, and only pay when you have something worth keeping online past 30 days. The free tier is a real trial of the whole workflow, not a demo.

The classic editor: when to write code yourself

Agent is the fast path, but Replit is still a code editor, and there are times to use it directly. Use Agent for architecture, scaffolding, and new features. When Agent starts making things worse, or over-engineering a small change, switch to Replit's Assistant, the inline autocomplete tool, for small surgical fixes. One experienced user quoted in the community guides put it this way: it is worth building the app twice, once to learn what is hard and how to prompt, and once to get it built right.

The editor is also where you learn what Agent actually built. Open the files, read the code, and change one small thing by hand. You do not need to understand every line, but knowing where the pieces live makes your next prompt better, because you can point Agent at a specific file instead of describing the whole app again. Our VS Code beginner guide covers the editor habits that transfer here: files on the left, code in the center, and a terminal for running commands.

Replit supports Git and GitHub integration, so you can push your project to a repository and keep a copy outside Replit. That matters because of hosting lock-in: your app lives on Replit's infrastructure, and there is no easy export to another host. Keeping the code in GitHub is the simple insurance policy. Our GitHub beginner guide walks through the first repository setup.

7 mistakes beginners make on Replit

Most first builds fail the same few ways. Each mistake below is easy to avoid once you know it, and all seven come from the guides and documentation fetched for this article.

1. One giant prompt. Asking for a complete platform in a single message produces a big, half-working app. Build in stages: core first, then accounts, then design, then integrations.
2. No boundaries. A prompt without a "do not build yet" list invites logins, payments, and databases into a first build that does not need them. Name the exclusions every time.
3. Skipping Plan mode. Building without a plan means discovering misunderstandings after the code exists. Review the task list before Agent writes anything.
4. Pasting secrets into chat. Never paste API keys, passwords, or personal records into Agent chat or any online tool. Use Replit's Secrets feature for credentials, and keep them out of prompts and screenshots.
5. Testing only on desktop. Most visitors will open your app on a phone. Resize Preview to a phone-sized width during the build, not after you publish.
6. Ignoring checkpoints. Agent creates a checkpoint at every step. Before a risky change, such as adding authentication or changing the database, make sure you have a clean checkpoint to roll back to.
7. Expecting a production app. Agent 4 is exceptional for prototyping, internal tools, and MVPs. For large production applications with complex state and business logic, professional developers still prefer local tools such as Cursor. Our Cursor AI beginner guide covers that side of the spectrum.

Replit vs Lovable vs Bolt vs Cursor

Replit is one of several AI app builders, and the right one depends on what you are building. All four below have beginner guides on this site, so you can compare the first-app flow for each before you commit to one.

The honest split is this: Replit, Lovable, and Bolt build apps for you, while Cursor helps you write code yourself. If you cannot code yet, start with Replit or one of the other builders. If you can code, or want to learn, Cursor is the tool that teaches you the code as you go. None of them is the right choice for a large production system on day one; that is a later decision, after the prototype works.

Sources

• Replit Docs — Replit Agent overview, updated 14 October 2026 — Agent definition, Plan mode, Free/Power/Max modes, five-hour allowance resets, Allow once confirmation, checkpoints, Design Canvas, Core 1 vs Pro 10 background tasks.
• Replit — Pricing and Plans, checked 28 October 2026 — Core $20 a month ($18 billed annually), Pro $100 a month ($90 billed annually), Free Mode hours and project allowances, Pro parallel agents and collaboration limits.
• Lilach Bullock — How to Use Replit: Build Your First App Step by Step, published and updated 23 October 2026 — the seven-step first-app flow, prompt boundaries, the 30-day free published link rule, badge removal, Starter vs Core table.
• Ray Fu — How to Use Replit's Agent 4: The Complete Beginner's Guide — Agent 4 launch in March 2026, infinite Design Canvas, parallel agents, multi-artifact projects, prompting rules, replit.md and design.md habits, honest limits.

Frequently Asked Questions

Is Replit free to use?

Yes. The Starter plan is free and includes daily Agent credits up to a monthly cap, monthly cloud credits, and one free published app. You can learn the whole workflow and ship a first app without paying. Core costs $20 a month, or $18 billed annually, when you want the app to stay online and the badge removed.

How long does a free Replit app stay online?

A published link on the free Starter plan automatically goes down after 30 days, according to Replit's documentation. That makes Starter right for learning and testing ideas, but not for a page your customers rely on. Move to Core before you send customers to a Replit link.

Do I need to know how to code to use Replit?

No. Agent handles the technical work: writing code, setting up infrastructure, and configuring databases. You describe what you want in plain language and Agent builds it. Knowing a little code helps when you read what Agent built or make small fixes in the editor, but it is not required to ship a first app.

What is Plan mode in Replit?

Plan mode is Agent's planning step. Agent breaks your project into an ordered task list, explores different approaches, and lets you review and refine the plan before any code is written. You enable it by clicking Plan in the chat input or asking Agent to enter Plan mode, and you approve the plan before building starts.

What is Free Mode in Replit?

Free Mode is no-cost everyday Agent work within your plan's allowance, with Replit choosing the AI model for you. On Core and Pro the allowances reset every five hours and have weekly limits. If a request needs more capability than Free Mode, Agent may offer a paid Allow once option, asks for confirmation first, and returns to Free Mode afterward.