Awesome Vibe Coding

🚀 The Vibe Coding Handbook

[](https://awesome.re) [](http://creativecommons.org/publicdomain/zero/1.0/) [](http://makeapullrequest.com)

The definitive handbook for the next era of software development. Vibe coding simplifies development through AI agents, enabling builders to ship highly complex products rapidly.

The term was coined by Andrej Karpathy in February 2025:

> "There's a new kind of coding I call 'vibe coding', where you fully give in to the vibes, embrace exponentials, and forget that the code even exists."

If you find this useful, please star this repo - it helps others discover it!

> Note from the maintainer: I built this handbook because I was drowning in AI coding tool launches and couldn't find a single place that had everything organized with honest, actionable advice. Whether you're a senior dev evaluating agents or a complete beginner who just heard about "vibe coding" - this is the resource I wish existed when I started. I personally test and verify tools before adding them. If something is missing or outdated, open an issue or submit a PR.

---

🧭 What Should I Use? (The Complete 2026 Decision Guide)

The AI coding landscape shifted radically in early 2026 (the "Agentic Engineering" era). It's no longer about fancy autocomplete; it's about composing stacks of autonomous software agents.

Don't just pick one tool. Pick your workflow.

🟢 Level 1: "I Have Zero Coding Experience"

You want to describe an app in plain English and get a working product—without ever touching a terminal.

• 🎯 For Full-Stack Web Apps: 👉 Use Lovable
• Why it wins: Lovable auto-generates your Next.js frontend, connects it to a real Supabase backend PostgreSQL database, and handles one-click deployments to platforms like Vercel.
• The "Vibe" Setup: Create an account, connect GitHub, type: "Build a habit tracker with user auth."
• Limitations: Struggles with complex, non-standard backend logic (e.g., custom webhooks or heavy cron jobs).
• Cost Thresholds: Free to start, but you hit message limits quickly. Pro is $20/mo.

• 🎨 For Landing Pages & UI Blocks: 👉 Use v0 or bolt.new
• Why it wins: v0 is unmatched for generating pixel-perfect React + Tailwind designs. Bolt.new goes further by running a full Node.js environment in your browser via WebContainers.
• Quick Start: Go to v0.dev. Prompt: "A dark-mode pricing table with glassmorphism."

---

🟡 Level 2: "I Want An AI-Native IDE"

You are a developer who wants visual feedback, a rich ecosystem, and AI deeply integrated into every keystroke.

• 👑 The Industry Standard: 👉 Use Cursor
• Why it wins: Cursor's "Composer" feature orchestrates parallel AI agents to handle massive multi-file refactors. It sits on top of VS Code, so all your extensions work flawlessly.
• Setup: Download app → Log in → (Optional) Add your own Anthropic API key to bypass Pro limits.
• Pro Tip (Composability): Start your app in Lovable, push it to GitHub, then clone and open it in Cursor for advanced architectural changes.
• Cost Thresholds: Free tier is very restrictive (slow requests). Pro ($20/mo) is essential for daily drivers.

• ⚡ The Context-Aware Alternative: 👉 Use Windsurf
• Why it wins: "Cascade" offers deeper, more persistent codebase awareness out of the box without needing extensive .cursorrules configurations. Features a generous free tier.

---

🔴 Level 3: "I'm a Power User. Give me Terminal Agents."

You want maximum control, privacy, and the ability to pipe different AI models (like Claude, Gemini, or Local Models) into your workflow.

• 🧠 For Maximum Reasoning Power (API/CLI): 👉 Use Claude Code or OpenAI Codex
• Why Claude wins: Anthropic's terminal agent natively harnesses Claude's 1M+ token context window. Best-in-class for deeply complex debugging.
• Why Codex wins: The pioneer in generative coding models, heavily updated for 2026. OpenAI's direct Codex CLI integrations offer unmatched latency and deep logical puzzle-solving via o1/Codex reasoning architectures.
• Performance Context: As of April 2026, Claude Opus variants and OpenAI Codex (o1-tier) paired with agentic scaffolding dominate the SWE-bench Verified leaderboard (consistently hitting between 75% and 82%).
• Quick Start: npm install -g @anthropic-ai/claude-code → claude
• Cost Thresholds: Pay-per-token API pricing. Warning: Extensive context use can rack up a $10 bill in a few hours if you aren't careful.

• 🔧 For TUI Workflows & Open Source: 👉 Use OpenCode or Aider
• Why OpenCode: Exploded in popularity in 2026. A powerful Terminal UI (TUI) that supports 75+ LLMs natively (use whatever API key you want).
• Why Aider: The veteran open-source tool with the most battle-tested, git-aware, multi-file editing capabilities.
• Quick Start (Aider): pip install aider-chat → aider --model claude-3-5-sonnet

---

🟣 Level 4: "I Want an Autonomous AI Software Engineer"

You want to assign a Jira ticket to an AI and go get coffee. Benchmark: Handling end-to-end multi-file implementations autonomously.

• 🤖 For Enterprise-Grade Autonomy: 👉 Use Devin
• Why it wins: The first fully sandboxed autonomous SWE. Incredible capabilities for handling full DevOps pipelines and deployments independently.
• Cost: Enterprise pricing.

• ✨ For Advanced Concurrent Agentic Power: 👉 Use Google Antigravity
• Why it wins: Designed by Google DeepMind, Antigravity introduces revolutionary "Planning Mode" paradigms combined with extreme parallel tool execution. It acts less like a simple code editor and more like a Chief Technology Officer managing a local multi-agent system.
• Performance Benchmark: Antigravity excels in navigating massive monorepos by dynamically creating multi-step implementation plans and requesting human architectural feedback before mutating code.

• 🔓 The Open Source Champion: 👉 Use OpenHands
• Why it wins: The premier free, self-hosted alternative to Devin. Run it locally via Docker, hook it up to Claude's API, and let it resolve your GitHub PRs autonomously.
• Performance Benchmark: OpenHands + Claude scores incredibly high because the sophisticated agentic scaffolding allows the LLM to run bash commands and test its own code before submitting.
• Quick Start (Docker required):

``bash docker run -it --pull=always \ -e WORKSPACE_BASE=$(pwd)/workspace \ -v /var/run/docker.sock:/var/run/docker.sock \ -p 3000:3000 docker.all-hands.dev/all-hands-ai/openhands:0.8 ``

---

🥊 Deep Dives & Architectural Comparisons

If you are evaluating tools for an engineering team, don't rely on generic feature matrices. We have built deep technical head-to-head comparisons detailing workflow philosophies, actual SWE-bench data, and vendor lock-in risks for April 2026:

1. [The AI-Native IDE Battle: Cursor vs. Windsurf](comparisons/cursor-vs-windsurf.md)
2. [Terminal Agents: Claude Code vs. OpenCode (vs. Aider)](comparisons/claude-code-vs-opencode.md)
3. [App Builders: Lovable vs. Bolt.new vs. v0](comparisons/lovable-vs-bolt-vs-v0.md)
4. [Autonomous SWEs: Devin vs. OpenHands](comparisons/devin-vs-openhands.md)
5. [System Architectures: Google Antigravity vs. OpenAI Codex](comparisons/antigravity-vs-codex.md)

---

⚡ Instant Value: Copy/Paste Templates

Don't start from scratch. Use our April 2026 standardized templates to instantly upgrade how your AI coding tools behave. Drop these into your repositories right now:

• 🛠️ [Advanced .cursorrules Template](templates/cursorrules-advanced.md) (TypeScript/Next.js focused system instructions)
• 🖥️ [Terminal Agent CLAUDE.md Template](templates/claude-system-prompt.md) (Safety and TDD rules for CLI tools like Claude Code and Aider)
• 📋 [Internal Team "Vibe Guide"](templates/vibe-guide.md) (An Engineering Manager's policy for teams using AI)

🍳 The Prompt Recipes Cookbook

Context Engineering is the hardest part of AI development. Don't write prompts from scratch; use our framework-agnostic snippets:

• 📝 [01. The New Project Checklist](prompts/01-new-project-checklist.md)
• 🏗️ [02. Architecture Scaffolding Prompts](prompts/02-architecture-scaffolding.md)
• 🐛 [03. Debugging & Refactoring Prompts](prompts/03-debugging-and-refactoring.md)

📦 Vibe Coding Starter Kits

Need to instantiate a perfect environment? Use our setup guides:

• ✨ [Google Antigravity Elite Terminal Setup](starters/antigravity-terminal-stack.md) (For massive concurrent logic tasks)
• 🌐 [The Universal Web Stack](starters/universal-web-stack.md) (For standard Cursor / Next.js web applications)

---

🚧 How to Not Fail: Common Mistakes

AI coding tools are force multipliers, meaning they multiply your mistakes just as fast as your successes. Read our Deep Dive guides so you don't fall into the same expensive traps as everyone else:

• ❌ [5 Mistakes Everyone Makes with Cursor (and How to Fix Them)](guides/mistakes-cursor.md)
• 💸 [5 Mistakes Everyone Makes with Claude Code & Terminal Agents](guides/mistakes-claude-code.md)

---

Contents

• [AI Coding Agents](#ai-coding-agents)
• [IDE-Based](#ide-based)
• [Terminal / CLI](#terminal--cli)
• [Browser-Based](#browser-based)
• [Autonomous Software Engineers](#autonomous-software-engineers)
• [AI App Builders (No-Code)](#ai-app-builders-no-code)
• [AI Code Completion & Assistants](#ai-code-completion--assistants)
• [AI Code Review](#ai-code-review)
• [AI for Design → Code](#ai-for-design--code)
• [AI Debugging & Testing](#ai-debugging--testing)
• [AI Documentation Tools](#ai-documentation-tools)
• [Prompt Engineering for Code](#prompt-engineering-for-code)
• [Guides & Techniques](#guides--techniques)
• [Prompt Libraries & Templates](#prompt-libraries--templates)
• [Rules Files & System Prompts](#rules-files--system-prompts)
• [Context Engineering](#context-engineering)
• [MCP Servers for Coding](#mcp-servers-for-coding)
• [Official MCP Servers](#official-mcp-servers)
• [Developer Tools](#developer-tools)
• [Database](#database)
• [Productivity & Business](#productivity--business)
• [Web & Browser](#web--browser)
• [Design](#design)
• [Search](#search)
• [MCP Directories & Aggregators](#mcp-directories--aggregators)
• [AI Workflow Automation](#ai-workflow-automation)
• [Multi-Agent Frameworks](#multi-agent-frameworks)
• [Local AI for Coding](#local-ai-for-coding)
• [Local Inference Engines](#local-inference-engines)
• [Open Source Code Models](#open-source-code-models)
• [AI-Powered DevOps & Infrastructure](#ai-powered-devops--infrastructure)
• [AI for Mobile Development](#ai-for-mobile-development)
• [AI for Game Development](#ai-for-game-development)
• [AI Git & Version Control](#ai-git--version-control)
• [AI Browser Extensions](#ai-browser-extensions)
• [AI for Data Science & Analytics](#ai-for-data-science--analytics)
• [AI-Powered Backend & APIs](#ai-powered-backend--apis)
• [AI Image Generation](#ai-image-generation)
• [AI Video Generation](#ai-video-generation)
• [AI Audio & Music](#ai-audio--music)
• [AI Writing Tools for Developers](#ai-writing-tools-for-developers)
• [AI Search & Research](#ai-search--research)
• [Benchmarks & Comparisons](#benchmarks--comparisons)
• [Learning Resources](#learning-resources)
• [Articles & Blog Posts](#articles--blog-posts)
• [YouTube Channels & Videos](#youtube-channels--videos)
• [Courses](#courses)
• [Books](#books)
• [Newsletters](#newsletters)
• [Podcasts](#podcasts)
• [Communities](#communities)
• [Case Studies & Showcases](#case-studies--showcases)
• [Best Practices & Workflows](#best-practices--workflows)
• [Opinions & Criticism](#opinions--criticism)
• [Related Awesome Lists](#related-awesome-lists)

---

AI Coding Agents

IDE-Based

Full-featured code editors with deeply integrated AI that can understand your entire codebase, edit multiple files, and execute commands autonomously.

• Cursor - AI-first code editor built on VS Code. Industry benchmark for agentic IDE experience. Features Composer for multi-file editing, Tab for intelligent autocomplete, and deep codebase awareness. Personal pick - the one most people should start with. [Freemium] [Pro $20/Mo]
• Windsurf - AI coding IDE by Codeium (acquired by Cognition in 2025). Strong context awareness across large codebases at competitive pricing. Features Cascade for multi-step agentic flows. [Freemium]
• Void - Open-source AI code editor. Privacy-first alternative that lets you bring your own model/API key. No vendor lock-in. [Free] [Open Source]
• Zed - High-performance editor written in Rust with built-in AI assistant. Known for blazing speed and multiplayer editing. [Free]
• PearAI - Open-source AI code editor forked from VS Code with integrated AI chat and inline editing. Community-driven. [Free] [Open Source]
• Melty - Open-source AI code editor that watches what you do and understands your codebase changes in real-time. [Free] [Open Source]
• Trae - AI-native IDE by ByteDance. Free access to Claude and GPT models with agentic Builder mode. [Free]
• VS Code + Copilot - Microsoft's editor with GitHub Copilot chat, inline suggestions, and agent mode deeply integrated. [Copilot Free Tier Available]
• JetBrains IDEs + AI Assistant - IntelliJ, PyCharm, WebStorm etc. with JetBrains' own AI assistant and Copilot support. [Paid]
• Neovim + AI Plugins - The hackable text editor with AI plugins like Copilot.vim, ChatGPT.nvim, and Avante.nvim for terminal purists. [Free] [Open Source]
• Emacs + GPTel - AI assistant for Emacs. Chat with LLMs, generate code, inline editing. For Emacs die-hards. [Free] [Open Source]
• Fleet - JetBrains' lightweight editor with AI capabilities. Polyglot, fast, collaborative. [Free]

Terminal / CLI

AI agents that operate from your terminal, reading, writing, and executing code with full system access.

• Claude Code - Anthropic's terminal-native agentic coding tool. Best-in-class reasoning with Claude Opus/Sonnet. Can read/write files, run commands, search codebases, and manage git. Personal pick - best for complex, multi-file tasks. [Requires Claude Max Subscription]
• Codex CLI - OpenAI's open-source CLI coding agent. Lightweight, fast, runs locally with configurable autonomy levels. [Free] [Open Source]
• Aider - Open-source AI pair programming in your terminal. Git-aware, works with any LLM. One of the original CLI coding tools. Best-in-class for its edit format system. Personal pick - best open-source option. [Free] [Open Source]
• Gemini CLI - Google's open-source terminal AI agent. Integrates with Gemini models with 1M token context. [Free] [Open Source]
• [Amazon Q Developer CLI](http

…

Source & license

This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: bluegalaxy111
• Source: bluegalaxy111/awesome-vibe-coding
• License: CC0-1.0

Install and usage instructions live in the source repository linked above.