The full-stack builder for vibe coders, product thinkers, and AI-native creators — all inside Cursor.
Cursor Project Master (CPM) transforms structured documents directly into full-stack, production-ready applications. Built specifically for Cursor users, CPM requires only a single clear Product Requirements Document (PRD) to generate, test, and deploy entire applications automatically.
- 📝 Just your docs: Clearly describe your idea—CPM takes care of everything else.
- 📋 Visual Kanban: Track tasks automatically from
todotodone. - 🤖 Zero hassle: Fully automated coding, testing, and deployment.
- 🚀 Instant launch: Move swiftly from idea to deployed product, iterate effortlessly.
You have two simple options to get started:
Option 1: Let Cursor Do It (Recommended) Open Cursor, start a new chat in project directory, and give it this instruction:
Initialize the project from
https://github.com/heyzgj/cursor-project-masterand set it up for me.
Option 2: Manual Install If you prefer the command line and have Node.js v18+ installed, run:
npx @project-master-ai/cli initOnce the files are in your project folder, open it in Cursor. In the AI chat, simply type:
/initThe AI will now guide you through setting up your project requirements. There is no need to manually edit any files. Just talk to the agent.
Cursor will fill in the templates located in docs/_templates/:
PRD.md– Product vision & requirementsTECH_SPEC.md– Technical stack & deployment preferencesDATA_MAP.md– Define your data clearlyUX_FLOW.md– User interactions step-by-stepSTYLE_GUIDE.md– Visual and UI elements
Tip: If you don't want to spend Cursor request, you can use ChatGPT or Gemini as assistants for clarity (prompt example below):
Act as a world-class product manager and a principal software architect. Your task is to help me create the initial documentation blueprint for a new software project. You must be rigorous, structured, and ask clarifying questions if my idea is ambiguous. ## CONTEXT You will be generating the content for a set of specific Markdown templates that an autonomous AI agent will use to build the entire application. The agent requires extreme clarity and structure. It cannot make assumptions. ## MY IDEA [Your IDEA] ## YOUR TASK Based on my idea, generate the content for the following files(PRD/TECH SPEC/...). For each file, present the content inside a Markdown code block. >**Critical Rules**>> 1. Do **not** invent, infer, or guess any information. If something is unknown, ask. > 2. Before drafting **any** document you must: > a. Ask every clarifying question needed for 100 % accuracy. > b. Wait until I reply **"done"** to signal that no more information will be provided for now. > c. Request that I upload the blank template file you're about to fill.> 3. Work **one file at a time** in the order above. Do not begin the next file until I approve the current one.> 4. When writing:> • Use precise, imperative language (avoid words like "robust," "nice," "etc.").> • If a detail is truly undecided, write **"N/A – not decided yet"** instead of leaving it blank or making assumptions.> • Output pure Markdown **without code-fences** so I can paste directly into the file.> 5. After you deliver a draft, explicitly ask: **"Is this acceptable? If yes, reply 'next'; if changes are needed, specify them."**> **Workflow Overview**>> 1. **Discovery Phase** – You ask exhaustive questions covering product idea, personas, constraints, metrics, stack, data, UX, style, compliance, etc.> 2. I answer until satisfied, then type **"done."**> 3. You say: *"Please upload `PRD.md` template."*> 4. I upload the file.> 5. **Drafting Phase** – You generate the completed `PRD.md` content only.> 6. I review; if approved, I reply **"next."**> 7. Repeat steps 3-6 for `TECH_SPEC.md`, then `DATA_MAP.md`, `UX_FLOW.md`, and finally `STYLE_GUIDE.md`.> 8. After the fifth file is accepted, you conclude with **"All documents completed."**> **Begin by asking your clarifying questions now.**Place completed docs in /docs.
- Tasks auto-update through stages (
todo,in_progress,done). - Track live updates in
project/project_status.md. - Visual Kanban interface under
/kanbanfor real-time tracking. - Auto-deploy preview on every commit; production deployment by merging into
main.
# Start Kanban board directly (in any CPM project) npx @project-master-ai/cli kanban # Or with custom port npx @project-master-ai/cli kanban --port 3001The Kanban board will automatically open in your browser!
Edit your PRD:(Example)
## New Feature: Image Upload Allow users to upload images (JPEG/PNG ≤ 5MB). Store on Supabase Storage.Then just inform CPM:
Add the "Image Upload" feature from PRD.md. CPM takes care of updating, implementing, testing, and deploying automatically.
.cursor/rules/ # CPM guidelines kanban/ # AI-powered Kanban interface docs/ # Product & technical specs project/ _templates/ # Task templates tasks/ # Kanban-style management project_status.md # Real-time updates scripts/ # Utilities for reporting .github/workflows/ # CI/CD automation src/ # Auto-generated code | Question | Answer |
|---|---|
| Am I stuck with one tech stack? | No. Edit TECH_SPEC.md and CPM adapts seamlessly. |
| Can I manually edit the code? | Yes! Your changes become the new baseline instantly. |
| How does CPM store "memory"? | Clearly in local text files. Nothing hidden. |
| What if tests fail repeatedly? | CPM reflects, revises plans, retries intelligently. |
MIT License – Fork, build, innovate!
Contributions welcome! PRs are encouraged—let's build together.