The vibe coding directory
46 tools across 7 categories. Every entry says what it's genuinely good at and where it bites, not just what the homepage claims.
Nothing matches that. Try a broader term.
AI Coding Agents
Tools that read your codebase, plan a change, and write the files themselves. This is the engine room of vibe coding — pick one and get fluent in it before you collect five.
Claude Code
Anthropic's agentic coding tool for terminal, desktop, web, and IDE.
Cursor
An AI-first code editor built on VS Code.
Windsurf
Agentic IDE with a strong codebase-awareness story.
Cline
Open-source autonomous coding agent for VS Code.
Aider
Terminal pair programmer that commits as it goes.
GitHub Copilot
The original inline completion, now with agent modes.
Continue
Open-source IDE assistant, any model you like.
Zed
A very fast editor with agentic editing built in.
OpenAI Codex
OpenAI's cloud-based software engineering agent.
Roo Code
Cline fork with specialized operating modes.
Prompt-to-App Builders
Describe an app, get a working app. Fastest path from idea to something clickable, and the most common place projects stall once the prompt stops being enough.
Models
The brains underneath the tools. Most agents let you swap models, and the choice changes cost, speed, and how far the agent gets before it needs you.
Claude Opus 5
Anthropic's most capable model.
Claude Sonnet 5
The balanced workhorse of the Claude 5 family.
Claude Haiku 4.5
Fast and cheap for high-volume work.
GPT-5 family
OpenAI's frontier line.
Google Gemini
Very large context windows and tight Google integration.
DeepSeek
Open-weight models with strong coding scores at low cost.
Qwen Coder
Alibaba's open-weight coding model line.
Ollama
Run open models locally with one command.
MCP & Integrations
Model Context Protocol lets an agent reach past your code into browsers, databases, APIs, and the live web. This is how you stop pasting things into chat.
Model Context Protocol
The open standard for connecting agents to tools and data.
Playwright MCP
Let the agent open a browser and check its own work.
GitHub MCP Server
Issues, pull requests, and repo operations as agent tools.
Supabase MCP
Let the agent inspect and query your database.
Filesystem MCP
Scoped file access for agents that don't otherwise have it.
Context7
Current library documentation, injected on demand.
Deploy & Backend
Where the thing actually lives. Vibe-coded frontends are easy; auth, databases, and hosting are where most builders hit the first real wall.
Netlify
Git-connected hosting for static sites and functions.
Vercel
The default home for Next.js.
Supabase
Postgres with auth, storage, and APIs attached.
Railway
Deploy backends and databases without writing infrastructure config.
Cloudflare Workers & Pages
Edge compute, storage, and hosting at low cost.
Neon
Serverless Postgres that branches like git.
UI & Design
Component systems and styling that AI models already know deeply — using them means the agent writes code it has seen a million times instead of inventing CSS.
Guardrails & Quality
Tests, error tracking, and review tooling. The difference between a demo and something you can leave running is almost entirely in this category.
Playwright
Browser testing that an agent can write and run itself.
Vitest
Fast unit testing for modern JavaScript projects.
Sentry
Find out about errors before your users tell you.
CodeRabbit
AI code review on every pull request.
Snyk
Find vulnerable dependencies and insecure patterns.
Dependabot
Automated dependency updates, built into GitHub.
Git
The undo button that makes agentic coding survivable.
Directory last reviewed 2026-08-15. Tooling in this space moves fast. Pricing tiers and model lineups change often — treat 'pricing' as a coarse shape, not a quote. Re-verify before publishing comparisons.