Two categories, not one ranking
Most comparisons of Claude Code, Cursor, Codex, Lovable, v0, and Bolt treat this as one big ranking — which tool is objectively best. That's the wrong question. These six split into two genuinely different categories, and which one you need depends entirely on what you're building, not on which one is trending this week.
The first category works inside a real, existing codebase: Claude Code, Cursor, and Codex read, understand, and edit files across a real project, the same way a developer would. The second category generates a working app from a prompt: Lovable, v0, and Bolt turn a description into a scaffolded, mostly-working product, hiding the code by default.
That split maps directly onto how every verdict on this register already thinks about effort. It's worth knowing which category you're actually choosing between before you pick a specific tool inside it.
When you need a real codebase: Claude Code, Cursor, Codex
These three read an entire project, understand how its pieces connect, and make multi-file changes the way a developer working directly in the code would. That's the right tool once there's a genuine data model, real backend logic, or an existing project to extend rather than start fresh — this register's own "a weekend"-and-up territory, not an evening's scaffold.
Within this category the real difference is interaction style, not capability. Claude Code is agent-first: you describe a task, it works across the codebase autonomously, and you review the result. Cursor is editor-first: you stay in the loop, reviewing each change as a diff before it lands, inside what's functionally a VS Code fork. Codex, OpenAI's entry, sits in similar territory to Claude Code — an agent you hand a task to rather than an editor you work inside.
One detail worth knowing before treating "Cursor vs Claude Code" as a clean either/or: Cursor is model-agnostic. It can route requests to Claude, GPT, Gemini, or others, which means you can get Claude's reasoning inside Cursor's interface. The real choice often isn't the model — it's whether you want to hand off a task or stay hands-on reviewing it as it happens.
When you need a fast, visual scaffold: Lovable, v0, Bolt
These generate a working app from a prompt rather than editing an existing project — closer to this register's "an evening" territory: landing pages, quick prototypes, validating an idea visually before committing more time to it.
They're not interchangeable with each other either. Lovable ships the most full-stack by default — auth, a Supabase database, and Stripe wired in without asking you to configure any of it, aimed at going from an idea to something a real user could pay for in a single sitting. v0 is the most Vercel-native, generating clean React and Next.js code with shadcn/ui components, and works best if that's already the stack you're targeting. Bolt is the most framework-flexible, running a full development environment inside the browser with support for React, Vue, Svelte, and even React Native, useful when you don't want to commit to one framework upfront.
One honest caveat worth taking seriously: multiple independent reviews of this category flag the same gap, between a demo that looks finished and a scaffold that's actually production-ready. Database policies, security review, and real backend maturity are frequently the part that still needs real attention before anything here goes in front of paying users — fast is genuinely fast, but it isn't the same thing as done.
How to actually decide for your idea
Start with the effort this register would already put on your idea. If it's genuinely an evening's build — a landing page, a simple form, something without real backend logic — a scaffold builder gets you there fastest, and Lovable specifically if you want auth and payments wired in without configuring them yourself.
If it's a weekend or bigger — a real data model, meaningful backend logic, something you're extending rather than generating fresh — go straight to an agentic tool working in a real codebase. Worth noting plainly: nothing in this register's 300-plus entries has needed a bucket bigger than that yet, which says something about how far a well-scoped solo build genuinely goes with either category.
The two categories aren't mutually exclusive within a single project, either. A common, sensible pattern is scaffolding the UI and validating the idea quickly in Lovable or v0, then moving into Claude Code or Cursor once the project needs real backend logic the scaffold builder doesn't cover. Pick the category for what the moment actually needs, not one tool for the whole build.