The Top 50 Vibecoded Projects Built With Claude Code and OpenAI Codex (2026 Edition)

July 6, 2026

$ vibe-coding --list-top-50

The Top 50 Vibecoded Projects Built With Claude Code and OpenAI Codex (2026 Edition)

Two years ago, "vibe coding" meant asking an AI to autocomplete a function. Now it means handing an agentic coding tool a raw idea and watching it plan, write, test, and ship the whole thing — games, internal tools, entire products — largely unsupervised. We pulled together the 50 most talked-about examples and project categories from the Claude Code and OpenAI Codex ecosystems, and linked every single one back to its source so you can go verify it yourself.

Methodology: every entry links to its original write-up, repo, or a live search you can browse — nothing here is unsourced.
Claude Code crossed $1B in annualized run-rate in 2026, and its agentic "plan, then build" workflow is behind most of the enterprise-scale examples here.
OpenAI Codex passed 3 million weekly users after its dedicated Codex models shipped, adding roughly a million users a month.
Every builder on this list started with a prompt, not a blank IDE — the common thread is agentic tools doing multi-step planning and execution, not just line completion.

01–06Enterprise & Scale-Proof Builds

These are the examples that answer the "but does it actually hold up at scale?" question — real companies putting agentic coding tools on production-critical work.

OpenAI's internal zero-manual-code product — application logic, tests, CI config, and docs all written by Codex; roughly a million lines of code and ~1,500 PRs merged by a three-engineer team. Source ↗
Harvey's legal-engineering workflow — Codex cut early iteration time 30–50%, freeing engineers for system design instead of boilerplate. Source ↗
Wonderful's core architecture work — Codex CLI has reportedly replaced every other agentic harness the team used for deep-reasoning technical work. Source ↗
Tobi Lütke's MRI viewer — the Shopify CEO used Claude Code to build a working cross-platform viewer for a locked hospital MRI file format in a single sitting. Source ↗
Pietro Schirano's DNA analysis tool — the MagicPath founder fed Claude Code a raw consumer DNA file; it spun up multiple sub-agents to analyze different regions of the genome in parallel. Source ↗
Deedy's Hermès spec-ad pipeline — a Menlo Ventures partner prompted Claude Code to script, voice (via ElevenLabs), source visuals (Google Veo 3), and stitch (ffmpeg) a full 30-second ad from scratch. Source ↗

Shipped something with Claude Code that actually surprised you?

That's exactly the kind of milestone builders are putting on a shelf instead of letting it disappear into a group chat.

Shop the Anthropic Token Award — $99

07–12Games & Interactive Builds

WotAI's arcade-classic set — ten playable browser remakes (Pac-Man, Tetris, Space Invaders, and seven more), vibe-coded with prompts over a single weekend. Source ↗
The three-hour physics game — a solo builder shipped a live, published browser game with physics, sound effects, sparks, and a conquest mechanic in one afternoon session. Source ↗
The limited-context platformer — a head-to-head test of coding agents found Claude Code holding an edge building a functional Python platformer under tight context limits. Source ↗
The one-afternoon word game — an indie builder documented vibe-coding a full chaotic word game with Claude and publishing it free the same day. Source ↗
An "Oregon Trail for podcasting" — Slate's tech team used Claude Code to prototype a narrative game based on their own show's format. Source ↗
Fully agent-edited video production — builders are now cutting and sequencing entire videos inside Claude Code rather than a traditional editing suite. Source ↗

13–19Tools Vibe Coders Built for Each Other

PeonPing — Tony Sheng's notification plugin that plays Warcraft III Peon voice lines when Claude Code finishes a task; picked up 1,200+ GitHub stars and 40+ community sound packs within days. Source ↗
awesome-claude-code — the community-maintained, 45,000+ star GitHub list of agents, statuslines, plugins, and workflows that's become the de facto directory for the ecosystem. Source ↗
claude-statusbar-style tools — a growing category of status-line utilities showing 5-hour/7-day rate-limit windows, context usage, and per-session cost. Browse ↗
Vibeyard-style desktop IDEs — wrapper apps that run multiple Claude Code sessions in a "swarm" grid with session inspection and sharing. Browse ↗
Claude Code MCP servers — custom Model Context Protocol servers that connect Claude Code to a team's own tools, APIs, and data sources. Source ↗
Retro 2D game tutorials — the space-shooter-style build that's become a standard "second project" for people learning agentic coding. Source ↗
Full-stack deploy tutorials — backend, auth, database, and deployment built end-to-end in a single Claude Code session. Source ↗

Every one of these started as a milestone worth marking

First user, first dollar, first "wait, it actually works" moment — the Token Award is built for exactly that.

Shop the OpenAI Token Award — $99 Shop the Play Button Plaque — from $23

20–50The Patterns Behind the Rest — Browse Real Examples of Each

Not every vibecoded project gets a viral write-up — most live quietly on Indie Hackers, Show HN, and #buildinpublic threads. Each category below links to a live GitHub search so you can browse real, current examples yourself rather than take our word for it:

$What This Actually Tells Us

The through-line across all 50 isn't the tech stack — it's speed. Projects that would have needed a contractor, a sprint, or a "someday" slot on a roadmap are now getting built and shipped in a weekend, sometimes an afternoon. That compression is exactly why we've argued that the AI token award trend exists at all: when the milestone itself is arriving faster than ever, the recognition for it needs to catch up. And for the builders on this list who happen to be engineers rather than founders, our guide to gifts for AI engineers covers exactly this scenario.