🤖 AI Products · $20/mo

Can I vibe code Claude?

Anthropic's AI assistant for writing, analysis, coding, and long documents.

🔴Not reallyest. one sitting (for the client)

Same story as any frontier assistant: the chat interface is a weekend at most, and the model is everything. Claude's value is a large language model trained by Anthropic on vast compute, plus features like long-context document handling, projects, and tool use layered on top. You can absolutely build a chat app that calls the Claude API — in fact that is the intended way to use it — but that is renting the intelligence, not rebuilding it. The model is the product, and the model is exactly the thing no prompt can conjure.

The one-shot prompt
Build a Claude-style chat client for my own use on top of the Claude API. Include streaming responses, markdown and code rendering, multi-turn history stored locally, file/document upload that is passed into the model's context, and a projects concept that groups chats with a shared system prompt and reference files. Add regenerate, stop, and a temperature control. Use Next.js App Router with the API key held server-side and sensible rate limiting. Make the README clear that this is a client around a rented frontier model — the intelligence is Anthropic's model accessed via API, not anything built here.
Paste into Claude Code, Codex, or your agent of choice — then steer. Learn the moves in Prompt Engineering.
What you build

A chat app wired to the Claude API with document upload, long-context handling, and saved projects — a client for a model you rent, not a model.

What you give up
  • The underlying model — trained on compute you do not have
  • Tuned long-context handling, tool use, and artifacts
  • Projects, memory, and the surrounding product features
  • The inference infrastructure serving it reliably at scale
Why people still pay

You are paying for the model, not the window it shows up in. Training a frontier LLM takes a research lab and a data-center budget, so people rent that intelligence through the app or the API — the one part of this you fundamentally cannot vibe-code.

modelinfrastructure
Related
Actually build it

A prompt is the first move, not the whole game. Course 01 teaches you to take a prompt like this one to a shipped, working app — reviewing, correcting, and steering the agent the whole way.

Start Course 01 →