AI assistants · checked 2026-10-19Can I vibecode Claude Pro?
KINDA· weekend project
The Claude window is ordinary software: a list of chats, streaming replies, file uploads and a model picker. An agent builds that over the Anthropic API in one sitting, and open tools such as Open WebUI and LibreChat already do it for free. What no prompt builds is the model, so you keep paying Anthropic, per token instead of per month; light and medium use usually lands under $20, while heavy daily use of the top models can pass it. The gaps are research, connectors, memory, the apps and the Claude Code allowance that Pro bundles.
$20/moprice
$240you'd save a year
one sittingbuild time
—people replaced it
What you need on hand.
DifficultyIntermediate · Some system setup or keys.
TimeAbout an hour for the chat app; research, memory and phone access are extra sessions
- Claude Code, Codex or CursorAny of them can run the prompt.
- Node.js 22better-sqlite3 compiles a native module, so build tools must be installed.
- An Anthropic API key with prepaid creditAPI billing is separate from a Claude subscription; the prompt adds a monthly cap inside the app.
- Ollama (optional)For free local models. Small ones run on 16 GB of RAM; the larger open models want 32 GB or more.
- Your Claude exportRequested under Settings > Privacy on the web; the emailed link expires after 24 hours.
Before you buildTry Claude Free first: Sonnet and Haiku with up to 5 projects cost nothing. Pro matters when you need Opus, research or Claude Code.
- 1. Install one agent: Claude Code, Codex, Cursor.
- 2. Make an empty folder and open it in the agent.
- 3. Paste the prompt below, press enter, and answer its questions as it builds.
The prompt
Build me a private chat app to replace Claude Pro. Requirements: - Node.js 22 + TypeScript: a Fastify server and a React UI built with Vite. `npm start` serves everything on http://127.0.0.1:4710. Single user, no accounts, no telemetry, local-first; listen on 127.0.0.1 only. - Two providers behind one interface: Anthropic through the official `@anthropic-ai/sdk` (ANTHROPIC_API_KEY and a comma-separated ANTHROPIC_MODELS list in .env), and every model already pulled in Ollama on localhost:11434. Pick a model per chat. - Chat: streaming replies with Markdown and code highlighting (shiki), a stop button, regenerate, and edit-a-message to branch the chat from that point. Mark the system prompt and long pasted files with `cache_control` so repeat turns are billed at the cached rate. - Storage: better-sqlite3 in ./data/chats.db (chats, messages, model, input, output and cached tokens); uploads in ./data/uploads/. Text files and PDFs go to the model as documents, images as images. Full-text search over every chat with SQLite FTS5. - Projects: a folder of chats that share one system prompt and a few pinned files. - Cost meter: prices per model in ./prices.json; show what each reply cost, with cached tokens at the cached rate, plus a monthly total, and refuse to send once MONTHLY_CAP_USD from .env is reached. - Import my Claude export: find the conversations JSON in the ZIP, print its structure and a chat count first, then add every chat with its title and date. Importing twice must not duplicate chats. - Out of scope: voice, image generation, research mode, connectors, Claude Code, phone apps. - README: setup, where the data lives, and a plain note that the models are still Anthropic's, so every message sent to them is billed per token.
Opens your agent with the prompt filled in, you press enter. Or copy it and paste it into an empty folder.
Then keep going
Next prompts, same folder.
The first prompt builds the core. Run these one at a time in the same folder to grow it into your own Claude Pro.
Extend the projects in the chat app in this folder. Requirements: - A project page with a system prompt box and a file list (text, Markdown, PDF); show the token count of each file and of the whole project. - Send project files with `cache_control` and show the share of input tokens served from cache under each reply. - Warn when pinned files exceed 60% of the model's context window. - Moving a chat into a project and out again keeps its history. - Update the README.
Add MCP tools to the chat app in this folder. Requirements: - A ./mcp.json file lists local stdio servers (command, args, env); connect with `@modelcontextprotocol/sdk` at startup and show each server's status on a settings page. - Offer the servers' tools to Claude as tool definitions; every tool call shows as a card with its arguments, and I click Allow or Deny before it runs. "Always allow" is per tool. - Cap tool rounds at 8 per reply and show the cost of the extra rounds. - Out of scope: remote servers and OAuth. - Update the README with two example servers.
Add a research mode to the chat app in this folder. Requirements: - A "Research" toggle asks for a budget in dollars (default 1.00), then loops: plan 3-6 sub-questions, search the web (Brave Search API key in .env), fetch pages and read them with @mozilla/readability, then write the report. - Stop when the budget is spent and say so in the report. - Every claim in the report links to a numbered source; list sources at the end. - Save the plan, the pages fetched and the cost in the chat so I can audit it. - Update the README with the Brave pricing link.
Add memory to the chat app in this folder. Requirements: - A memories table in ./data/chats.db: id, text, created, last used. - The model gets a save_memory tool. Every proposed memory shows as a card I approve or reject; nothing is saved without a click. - Approved memories go into the system prompt of new chats, newest first, capped at 1,500 tokens. - A settings page lists all memories with edit and delete, and a per-chat "no memory" switch for private chats. - Update the README.
Make the chat app in this folder usable from my phone. Requirements: - A responsive layout below 640px wide: chat list in a drawer, full-width input. - A web app manifest and icons so it installs to the home screen. - Reach it over Tailscale: the server keeps listening on 127.0.0.1 and the README explains `tailscale serve` to expose it to my devices only. - Never listen on 0.0.0.0 and add no login of its own. - Update the README with the Tailscale steps.
What you lose
- The Claude Code allowance inside the plan: on the API every token is billed
- Research, connectors, Claude in Chrome and the Microsoft 365 add-in
- Memory that carries across chats without you managing it
- The phone and desktop apps over one account
- A flat price: heavy daily use of the top models can cost more than $20 on the API
Why people still pay
For $20 people get Anthropic's strongest models plus Claude Code, research and connectors behind one login on every device, with a usage window instead of a per-message meter. Projects and chat history make leaving feel like starting over, and an export can't be imported into another personal Claude account.
Claude Pro or your own copy?
Over
Your usage
3h ×$/h
Claude Pro, 1 yr$240
Your own copy, 1 yr$185running + build session
You keep$55pays for itself in 1 months
Estimate · Local models through Ollama cost only electricity. On the API, Claude Sonnet 5.5 costs $2 in and $10 out per million tokens, so a typical reply is about a cent and 30 messages a day is roughly $9 a month. Opus 5.5 ($4 in, $20 out) roughly doubles that, and long files or Fable 5.1 ($10 in, $50 out) can push past $30; cached input is billed far lower. The build session is token usage on your agent plan.
Take your data, then cancel.
1 · Take your data out of Claude Pro
- 1On the web or in Claude Desktop, open Settings > Privacy and click Export data. The iOS and Android apps can't export.
- 2Claude emails a download link to the address on your account. You must be signed in to open it, and it expires 24 hours after delivery; repeat the export to get a new one.
- 3The export holds your conversation data and your account's user information. It can't be imported into another personal Claude account.
- 4Team and Enterprise members can export their own chats only after an owner switches that on, and uploaded file contents aren't included.
2 · Cancel Claude Pro
- 1On the web or in Claude Desktop: Settings > Billing, then Cancel.
- 2On iOS or Android, cancel from the Billing menu in the app, or through the App Store or Google Play if you've deleted the app.
- 3The cancellation takes effect at the end of the billing period and you keep the paid plan until then; cancel at least 24 hours before the next billing date.
- 4Refunds aren't automatic: open a Claude Refund Request under Get help. A refund request doesn't cancel the plan, so cancel as well.
- Export before you cancel and check that the chats you need are in the file; the pages we read don't say that project files are included, so copy those yourself.
- Annual billing is paid up front: check the renewal date before it comes around.
- App Store purchases are refunded only by Apple, at Apple's discretion.
- Monthly API credits on Max and Team can't be refunded, and credits you already hold stay usable until they expire.
The escape hatch
Don't feel like building it? These are already free.
Open WebUIFreeSelf-hosted chat for OpenAI-compatible APIs and Ollama, with document chat, web search, voice and user accounts. BSD-based license with a branding clause.★ 156k · active Oct 2026 · docker to self-host
WebVisit ↗LibreChatOpen sourceMIT chat interface for many providers at once, Claude included, with agents, MCP tools, a code interpreter and message search.★ 46k · active Oct 2026 · docker to self-host
WebVisit ↗JanOpen sourceDesktop app that downloads open models and runs them offline; it can also call cloud APIs with your key.★ 45k · active Oct 2026 · one-click install
Windows, macOS, LinuxVisit ↗
WebVisit ↗LibreChatOpen sourceMIT chat interface for many providers at once, Claude included, with agents, MCP tools, a code interpreter and message search.★ 46k · active Oct 2026 · docker to self-host
WebVisit ↗JanOpen sourceDesktop app that downloads open models and runs them offline; it can also call cloud APIs with your key.★ 45k · active Oct 2026 · one-click install
Windows, macOS, LinuxVisit ↗
The numbers
Claude Pro pricing
| Plan | Monthly | Annual, per month | What you get |
|---|---|---|---|
| Free | $0 | $0 | Sonnet and Haiku, everyday use inside rolling five-hour windows, up to 5 projects, memory and connectors. |
| Pro | $20 | $17 | At least 5x the Free usage per five-hour session; Opus, Sonnet and Haiku, with Fable through usage credits; Claude Code, research, projects, memory and connectors. |
| Max | $100 | — | 5x or 20x the Pro usage per session, Fable at 50% of weekly limits, and API credits included: $100 a month on 5x and $200 on 20x. |
| Team | $25 | $20 | Standard seats get more usage than Pro; Premium seats cost $125 monthly or $100 a month yearly and give 5x the Standard usage; up to $500 a month of pooled API credits. |
| Enterprise | — | $20 | Seat fee plus usage billed at API rates, with admin spend limits. |
- Free tier
- Usage is limited per rolling five-hour window and only Sonnet and Haiku are included; Claude Code and research are Pro features.
- Billing
- Pro bills monthly at $20, or yearly at $200 up front ($17 a month). Cancelling takes effect at the end of the billing period, so cancel at least 24 hours before the next billing date. Payments are non-refundable except where the Consumer Terms or the law say otherwise; eligible refunds go through a request to support.
- Hidden costs
- Usage is metered per five-hour window, and the newest model, Fable, reaches Pro only through paid usage credits. Heavy users who hit the Pro limits have one step up, Max, from $100 a month. Subscriptions bought in the iPhone or Android app are billed and refunded through Apple or Google.
- Source
- claude.com
Also on the list
Related apps
Free covers everyday chat on Sonnet and Haiku within rolling five-hour limits, with up to 5 projects, memory and connectors, but no Claude Code or research.
Kinda. The core is a weekend build with the prompt on this page, but real gaps remain: The Claude Code allowance inside the plan: on the API every token is billed, Research, connectors, Claude in Chrome and the Microsoft 365 add-in.
About $20 a month on the Pro plan, $240 a year (checked 2026-10-19).
The Claude Code allowance inside the plan: on the API every token is billed; Research, connectors, Claude in Chrome and the Microsoft 365 add-in; Memory that carries across chats without you managing it; The phone and desktop apps over one account; A flat price: heavy daily use of the top models can cost more than $20 on the API. If any of those matter to you, keep paying.
Yes: Open WebUI (free), LibreChat (open source), Jan (open source).