The short answer
You can make your first real video game with Claude in 2 to 12 hours of focused work, depending on which path you pick. The walkthroughs that work are the small, tightly-scoped ones: a one-screen catch game, a Breakout clone, a basic platformer. The ones that do not work are the open-ended ones ("a Minecraft, but better"). The skill you need is the patience to copy and paste, not the ability to write code.
If you have never coded: start with the no-install Phaser walkthrough (3 hours, nothing to install). If you are a kid (9 to 13): start with the kids edition. If you have abandoned a lot of projects and need a fast finish: the 2-hour Godot Breakout walkthrough. If you want a real weekend project: the 12-hour Unity 2D platformer. All four are free. All four end with a playable game uploaded to itch.io.
What "make a game with Claude" actually means
It does not mean typing "make me a game" into a chat box and getting a finished product back. It means using Claude as a fast pair-programming partner: you describe what you want in plain English, you copy the code Claude writes, you paste it into a file, you save the file, you open it in a browser (or run it in an engine), and you see your idea move on the screen. Then you iterate.
For a beginner, the difference between this and "old-school" game dev is enormous. Previously, the first 40 hours of trying to make a game went to syntax errors, configuration problems, and looking up which method to call. With Claude doing the typing, those 40 hours collapse to about 2 hours of pasting code into a file and watching it work. The barrier to actually shipping something has fallen further in the last 2 years than in the previous decade.
What has not changed: scope still kills projects. The "with Claude" part makes coding faster. It does not make designing a 200-hour game possible in a weekend. You still need to pick something small enough to finish. That part is on you.
What to expect (and not expect)
The honest version, so you do not give up at the first surprise:
- Claude is fast but not perfect. Roughly one in five code blocks will need a small fix or a follow-up clarification. The walkthroughs build this in: the universal fix when something looks wrong is to paste your whole file back to Claude and ask it to find the problem.
- You will hit Claude's free-tier message limit. The free Claude tier allows a finite number of messages per window of time. Most beginner walkthroughs finish inside the free tier comfortably, but if you run out mid-build, that is normal: wait, come back, the file is saved.
- You will not understand every line of code Claude returns. That is fine and expected for a first game. The walkthroughs explicitly tell you not to try to read the code. You learn what each piece does by changing it and seeing what changes on screen, not by parsing it line by line.
- Some things will trip you up that are not about code. Saving a file with the right extension, opening it in the right app, finding the developer console when something breaks. Beginner walkthroughs that take these seriously will have explicit "if this happens, do this" boxes for the common ones.
- The first finished game will feel both bigger and smaller than you expect. Bigger because: you made a real thing on a computer. Smaller because: it is a one-screen game with one mechanic. Both reactions are correct.
Pick your path
Four paths, ranked by activation energy (easiest first). All free, all end with a shipped game.
Path 1: Kids edition (ages 9-13)
If you are a kid, or you are setting up a child to try this: the kids walkthrough is the gentlest path. It uses TextEdit (on Mac) or Notepad (on Windows), a browser, and a free Claude account that an adult sets up. The kid builds a cozy star-catching game with no way to lose, in 3 short steps, in roughly an afternoon. The walkthrough has a dedicated "For grown-ups" panel covering the AI-account setup and safety. The finished game runs locally and there is an embedded playable version on the page so the kid can see the target.
Also useful: our parent's guide to game dev for kids covers what game dev for kids looks like generally, age-by-age tool recommendations, and the AI-supervision conversation.
Path 2: No-install browser game (3 hours)
If you have never coded but are not a kid: the no-install Phaser walkthrough is the same idea pitched at a curious adult. TextEdit (or Notepad), a browser, a Claude account, no Node.js, no terminal, no commands. You build a tiny arcade dodger in roughly 3 hours (or a full evening if it is the first time you have ever touched code). Ends with the game uploaded to itch.io as a shareable web link.
This is the path most beginners should take. It eliminates the install friction that kills more first-game attempts than any other single thing.
Path 3: 2-hour Breakout in Godot (the fast-finish path)
If you have started a lot of game projects and finished none of them, do not start a new ambitious one. Take the 2-hour Godot Breakout walkthrough. Godot is a free game engine with a small install. The walkthrough is four 30-minute steps that take you from empty project to a fully working brick-breaker game with score, lives, win and lose screens, exported to the web. Two hours, real finish line, your name on a shipped game. Then your brain remembers what finishing feels like, and the next finish is easier.
If "I never finish projects" describes you, read our piece on why solo devs never finish before you start, and consider using the scope-lock tool if your next project is something bigger.
Path 4: A real weekend project (12 hours)
If you are ready to commit a weekend and want something more substantial: the 12-hour Unity 2D platformer walkthrough. Unity is the most-shipped indie game engine, free for hobby use. The walkthrough takes you through a real platformer: coyote-time character controller, three tilemap levels, a patrolling enemy, coin pickups, sound, screen shake, a pause menu, a WebGL build uploaded to itch.io. About 12 hours of focused work split across two days. The walkthrough is honest about the time, with realistic per-step estimates that add up.
If you love retro pixel art and constraints: the Pico-8 puzzle walkthrough (~4 hours) is also excellent, but requires the one-time $15 Pico-8 purchase.
What you need, across any path
- A computer. Mac, Windows, or Linux. Not a tablet for the longer walkthroughs.
- A web browser. Anything modern (Safari, Chrome, Edge, Firefox).
- A free Claude account from claude.ai. The free tier is enough for beginner walkthroughs. Adults only; if you are setting this up for a child, the account should be in your name.
- A free itch.io account for shipping. Free, no credit card.
- Patience for copy and paste. The actual skill you need.
What you do not need: prior coding experience, paid software (Pico-8 is the only optional paid tool), expensive hardware, a Unity Pro license, an art degree, or a YouTube tutorial subscription.
Common pitfalls (and how to avoid them)
- Picking a project that is too big. The single biggest mistake. If you cannot finish your first game in a weekend, you have probably scoped wrong. See the finishing post for the full breakdown.
- Asking Claude for "a game" instead of for a specific small thing. "Make me a Mario clone" gets you nothing. "Make me a Phaser HTML file where a green square moves left and right with arrow keys" gets you something you can paste into a file and run. The walkthroughs include the prompts inline so you do not have to write them.
- Trying to read every line of code. Tempting, especially if you are technical. Skip it for the first game. Focus on what the game does on the screen and whether each step's check passes.
- Giving up when something breaks. Things will break. The universal fix is to paste your file back to Claude and ask Claude to find the problem. Treat broken code as a normal step in the process, not as a sign that you failed.
- Installing too many tools before you start. Every install is a stall. Start with the path that needs the fewest installs (the no-install Phaser walkthrough). You can always upgrade your toolchain later, once you know you like this.
After your first game
Finish the small one first. Then:
- Try the next walkthrough up. If you did no-install, try the 4-hour Phaser walkthrough or the 2-hour Godot Breakout. Each step up adds a little more toolchain in exchange for more capability.
- Pick a tiny original idea and lock the scope. Use the scope-lock tool to write down what your game is and is not, before you start. Then use Claude the same way the walkthroughs taught you.
- Use the prompts library for specific game-dev tasks you hit (coyote-time platformer movement, save/load, tile collision, particle effects). Each one is a tested, opinionated prompt that returns a single drop-in file. The short list of the prompts every solo dev reaches for: 5 Claude prompts every solo game dev should bookmark.
- Make some art. The sprite generator rolls a procedural pixel sprite in your browser. The sprite library has CC0 sprites you can use in any project.
- Lay out a level. The level builder generates a starter platformer or top-down level you can paint into your engine.
Why this is a good time to start
The gap between "wants to make a game" and "has shipped a small game" has never been smaller than it is today. Two years ago, a beginner without coding experience would spend their first 20 hours on syntax errors before seeing a player move on screen. With a co-dev like Claude, that 20 hours is more like 2. The rest of the work (designing, deciding what is in and out, finishing) is the same as it ever was, but you start much closer to the finish line.
The honest version of "make a game with AI" is not magic. It is a real shortcut through the boring parts so you spend your time on the parts that matter. Pick the smallest walkthrough that interests you, start this weekend, finish in an afternoon.