prd-writer-mcp 0.1.0

package/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 haandol
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

package/README.md

@@ -0,0 +1,208 @@
+# PRD Writer MCP Server
+
+A [Model Context Protocol (MCP)](https://modelcontextprotocol.io) server that helps you write PRD (Product Requirements Document) interactively with AI. Guides you through 9 structured sections with templates, conversation guides, and document management.
+
+## Features
+
+- 9-section PRD template with structured XML templates and conversation guides
+- Interactive Q&A workflow — AI asks focused questions, never auto-generates
+- Document management — create, save, load, and export as clean Markdown
+- Section dependency tracking — ensures referenced sections are reviewed first
+- Works with Claude Desktop, Cursor, Kiro, and any MCP-compatible client
+
+## Installation
+
+### Running with npx
+
+```bash
+npx -y prd-writer-mcp
+```
+
+### Manual Installation
+
+```bash
+npm install -g prd-writer-mcp
+```
+
+## Configuration
+
+### Claude Desktop
+
+Add to your `claude_desktop_config.json`:
+
+```json
+{
+  "mcpServers": {
+    "prd-writer": {
+      "command": "npx",
+      "args": ["-y", "prd-writer-mcp"]
+    }
+  }
+}
+```
+
+### Cursor
+
+1. Open Cursor Settings
+2. Go to Features > MCP Servers
+3. Click "+ Add new global MCP server"
+4. Enter the following:
+
+```json
+{
+  "mcpServers": {
+    "prd-writer": {
+      "command": "npx",
+      "args": ["-y", "prd-writer-mcp"]
+    }
+  }
+}
+```
+
+### Kiro
+
+1. Open the command palette (`Cmd + Shift + P` on Mac, `Ctrl + Shift + P` on Windows/Linux)
+2. Search for "MCP" and select **Kiro: Open user MCP config (JSON)** (or workspace-level)
+3. Add the following:
+
+```json
+{
+  "mcpServers": {
+    "prd-writer": {
+      "command": "npx",
+      "args": ["-y", "prd-writer-mcp"]
+    }
+  }
+}
+```
+
+Configuration file locations:
+
+- User level: `~/.kiro/settings/mcp.json`
+- Workspace level: `.kiro/settings/mcp.json`
+
+## Environment Variables
+
+| Variable         | Description                                                  | Default                   |
+| ---------------- | ------------------------------------------------------------ | ------------------------- |
+| `PRD_OUTPUT_DIR` | Directory for document files (`.prd.xml`, exported markdown) | Current working directory |
+
+Example with Claude Desktop:
+
+```json
+{
+  "mcpServers": {
+    "prd-writer": {
+      "command": "npx",
+      "args": ["-y", "prd-writer-mcp"],
+      "env": {
+        "PRD_OUTPUT_DIR": "~/Documents/prd"
+      }
+    }
+  }
+}
+```
+
+### Running from Source
+
+```bash
+git clone https://github.com/haandol/prd-writer.git
+cd prd-writer
+pnpm install
+pnpm build
+```
+
+Then configure your MCP client:
+
+```json
+{
+  "mcpServers": {
+    "prd-writer": {
+      "command": "node",
+      "args": ["/path/to/prd-writer/dist/index.js"]
+    }
+  }
+}
+```
+
+## Available Tools
+
+### Template Tools
+
+| Tool                    | Description                                           |
+| ----------------------- | ----------------------------------------------------- |
+| `get_prd_overview`      | Get the PRD template overview with conversation guide |
+| `list_prd_sections`     | List all available template sections                  |
+| `get_prd_section`       | Get a specific template section by number (1-9)       |
+| `get_prd_full_template` | Get the complete template with all sections           |
+| `get_prd_section_guide` | Get conversation guide for writing a section          |
+
+### Document Management Tools
+
+| Tool                      | Description                                 |
+| ------------------------- | ------------------------------------------- |
+| `init_prd_document`       | Create a new PRD document (`.prd.xml`)      |
+| `load_prd_document`       | Load an existing document to resume editing |
+| `save_prd_section`        | Save content to a specific subsection       |
+| `read_prd_section`        | Read current content of a section           |
+| `get_prd_document_status` | Get status of all sections                  |
+| `export_prd_markdown`     | Export as clean Markdown                    |
+
+## Workflow
+
+The server guides AI through a structured workflow:
+
+1. **Initialize** — `init_prd_document()` or `load_prd_document()`
+2. **Overview** — `get_prd_overview()` to get the conversation guide
+3. **For each section (1-9):**
+   - `get_prd_section_guide(N)` — get questions and criteria
+   - `get_prd_section(N)` — get the template structure
+   - Ask focused questions (1-2 at a time)
+   - `save_prd_section(N, ...)` — save after user confirmation
+4. **Export** — `export_prd_markdown()` for the final document
+
+## PRD Sections
+
+| #   | Section                     | Dependencies |
+| --- | --------------------------- | ------------ |
+| 1   | Overview                    | —            |
+| 2   | MVP Goals and Key Metrics   | —            |
+| 3   | Demo Scenario               | Section 2    |
+| 4   | High-Level Architecture     | —            |
+| 5   | Design Specification        | Section 6    |
+| 6   | Requirements Summary        | —            |
+| 7   | Feature-Level Specification | Section 6    |
+| 8   | MVP Metrics                 | Section 2, 6 |
+| 9   | Out of Scope                | —            |
+
+## Document Format
+
+Documents are stored in XML for reliable section parsing:
+
+```xml
+<prd-document project="My Project">
+
+<section id="1" title="Overview">
+<subsection id="1.1" title="Purpose">
+Content here...
+</subsection>
+</section>
+
+...
+</prd-document>
+```
+
+Export to clean Markdown via `export_prd_markdown()`.
+
+## Development
+
+```bash
+pnpm install
+pnpm dev        # Run with tsx (watch mode)
+pnpm build      # Build for production
+pnpm start      # Run built version
+```
+
+## License
+
+Apache-2.0

package/dist/constants.d.ts

@@ -0,0 +1,5 @@
+export declare const TEMPLATES_DIR: string;
+export declare const CHAPTERS_DIR: string;
+export declare const GUIDES_DIR: string;
+export declare const SECTION_TITLES: Record<number, string>;
+export declare const SECTION_REFERENCES: Record<number, number[]>;

package/dist/constants.js

@@ -0,0 +1,23 @@
+import path from "path";
+import { fileURLToPath } from "url";
+const __dirname = path.dirname(fileURLToPath(import.meta.url));
+export const TEMPLATES_DIR = path.join(__dirname, "templates");
+export const CHAPTERS_DIR = path.join(TEMPLATES_DIR, "chapters");
+export const GUIDES_DIR = path.join(__dirname, "guides");
+export const SECTION_TITLES = {
+    1: "Overview",
+    2: "MVP Goals and Key Metrics",
+    3: "Demo Scenario",
+    4: "High-Level Architecture",
+    5: "Design Specification",
+    6: "Requirements Summary",
+    7: "Feature-Level Specification",
+    8: "MVP Metrics",
+    9: "Out of Scope",
+};
+export const SECTION_REFERENCES = {
+    3: [2],
+    5: [6],
+    7: [6],
+    8: [2, 6],
+};

package/dist/guides/01.md

@@ -0,0 +1,27 @@
+<section_guide number="1" title="Overview">
+<purpose>Define product vision, target users, core problem, solution strategy, success criteria, and differentiators</purpose>
+
+<questions>
+1. What is the main purpose of the project?
+2. What is the official project name?
+3. Who are the target users?
+4. What is the core problem to solve?
+5. What is the solution strategy and key differentiator?
+</questions>
+
+<example>
+### 1.1 Purpose
+- Enable users to automatically generate technical/functional specification documents by conversing with AI
+- Help developers, product managers, and planners write MVP technical specs more efficiently
+
+### 1.2 Document Title
+
+- PRD(Agentic Lean Prototyping Spec) Writer
+
+### 1.3 Author
+
+- haandol <ldg55d@gmail.com>
+  </example>
+
+<completion>Print the full section after all items are written, then get user confirmation</completion>
+</section_guide>

package/dist/guides/02.md

@@ -0,0 +1,21 @@
+<section_guide number="2" title="MVP Goals and Key Metrics">
+<purpose>Define 2-5 measurable goals to validate the MVP hypothesis</purpose>
+
+<questions>
+1. What is the core hypothesis you want to validate with the MVP?
+2. Define 2-5 measurable goals to validate this hypothesis
+3. What are the baseline (current) and target values for each goal?
+</questions>
+
+<example>
+### 2.1 Purpose
+- Can AI quickly generate technical specification documents based on a given template, reflecting user input?
+
+### 2.2 Key Performance Indicators (KPIs)
+
+- Average time for a user to complete a document via AI: **under 30 minutes**
+- Conversation context retention rate: **90% or higher**
+  </example>
+
+<completion>Confirm after writing goals with quantitative metrics</completion>
+</section_guide>

package/dist/guides/03.md

@@ -0,0 +1,16 @@
+<section_guide number="3" title="Demo Scenario" references="2">
+<purpose>Write a demo scenario that validates the core hypothesis</purpose>
+
+<required_review>
+MUST review Section 2 (MVP Goals) before writing this section.
+Call read_prd_section(2) and summarize key goals before proceeding.
+</required_review>
+
+<questions>
+1. How can you demonstrate the goals from Section 2?
+2. What are the starting and ending points of the demo?
+3. What is the key user journey?
+</questions>
+
+<completion required="true">Confirm after writing the scenario aligned with Section 2</completion>
+</section_guide>

package/dist/guides/04.md

@@ -0,0 +1,31 @@
+<section_guide number="4" title="High-Level Architecture">
+<purpose>Describe the system architecture using C4 model Context and Container diagrams</purpose>
+
+<questions>
+1. What are the main components of the system?
+2. Are there any external system/service integrations?
+3. What is the rationale for the technology stack choices?
+</questions>
+
+<example>
+### 4.1 System Diagram
+```mermaid
+flowchart LR
+    User -- "Chat with AI" --> Chainlit(Python App)
+    Chainlit -- "LLM Query" --> Bedrock(Claude Sonnet 4.6)
+    Chainlit -- "Web Search" --> Tavily(Tavily API)
+    Chainlit -- "Save File" --> FileSystem(Local)
+```
+
+### 4.2 Technology Stack
+
+| Component          | Technology                |
+| ------------------ | ------------------------- |
+| Frontend & Backend | Python (Chainlit)         |
+| LLM API            | Amazon Bedrock, Langchain |
+| Web Search API     | Tavily API                |
+
+</example>
+
+<completion>Include Context/Container diagram descriptions</completion>
+</section_guide>

package/dist/guides/05.md

@@ -0,0 +1,36 @@
+<section_guide number="5" title="Design Specification" references="6">
+<purpose>Detail the UX, page flow, key screens, and user journey</purpose>
+
+<required_review>
+MUST review Section 6 (Requirements Summary) before writing this section.
+Call read_prd_section(6) and list Feature IDs (F1, F2...) to use in Key Pages.
+</required_review>
+
+<questions>
+1. How many key screens (pages) are there?
+2. What is the core functionality of each screen? (Use Feature IDs from Section 6)
+3. What is the navigation flow between screens?
+</questions>
+
+<example>
+### 5.1 Key Screens
+- **Chat Interface**: Main screen where users converse with AI to write documents (F1, F2)
+
+### 5.2 User Flow
+
+```mermaid
+sequenceDiagram
+    participant User
+    participant AI (Chainlit)
+    participant Claude (LLM)
+
+    User->>AI (Chainlit): "Start new document"
+    AI (Chainlit)->>Claude (LLM): Request template-based question generation
+    Claude (LLM)-->>AI (Chainlit): Question response
+    AI (Chainlit)-->>User: Display question
+```
+
+</example>
+
+<completion>Define key screens and flow (including Feature ID mapping)</completion>
+</section_guide>

package/dist/guides/06.md

@@ -0,0 +1,45 @@
+<section_guide number="6" title="Requirements Summary">
+<purpose>Enumerate functional/non-functional requirements and assign priorities</purpose>
+
+<questions>
+1. List the core functional requirements
+2. What is the priority of each requirement? (Must-Have / Should-Have / Nice-to-Have)
+3. What are the non-functional requirements? (up to 3)
+</questions>
+
+<workflow>
+After completing 6.1 and 6.2:
+1. Analyze the features defined in 6.1 and automatically infer dependency relationships between them
+2. Generate a Mermaid dependency diagram (graph TD) for Section 6.3
+3. Present the generated diagram to the user and ask for confirmation or corrections
+4. Do NOT ask the user to describe dependencies from scratch — propose your analysis first
+</workflow>
+
+<example>
+### 6.1 Functional Requirements
+- **F1: Template-based AI Document Writing** - AI generates documents based on a predefined PRD template, reflecting user input
+- **F2: Conversational Input Support** - Users can fill in the document step by step through conversation with AI
+- **F3: Document Download** - Download the completed document in Markdown format
+
+### 6.2 Non-Functional Requirements
+
+- **NF1: Performance** - Support 1,000 concurrent users, document generation within 30 seconds
+- **NF2: Security** - OAuth authentication via AWS Cognito
+
+### 6.3 Feature Dependency Diagram
+
+```mermaid
+graph TD
+    F1[F1: Template-based AI Document Writing]
+    F2[F2: Conversational Input Support]
+    F3[F3: Document Download]
+
+    F2 -->|depends on| F1
+    F3 -->|depends on| F1
+```
+
+</example>
+
+<important>Assign a unique ID to each functional requirement (F1, F2, ...)</important>
+<completion required="true">Confirm after all requirements have been assigned IDs</completion>
+</section_guide>

package/dist/guides/07.md

@@ -0,0 +1,56 @@
+<section_guide number="7" title="Feature-Level Specification" references="6">
+<purpose>Write detailed user stories for each requirement from Section 6</purpose>
+
+<required_review>
+MUST review Section 6 (Requirements Summary) before writing this section.
+Call read_prd_section(6) and confirm all Feature IDs (F1, F2...) to map 1:1.
+</required_review>
+
+<workflow>
+CRITICAL: Each Feature (7.x) must be written and saved individually.
+
+1. Confirm the Feature list from Section 6 (F1, F2, F3...)
+2. Proceed sequentially for each Feature:
+   a. Discuss the Feature using the questions below
+   b. Complete the 7.x subsection
+   c. Get user confirmation
+   d. Call save_prd_section(7, content, subsection=x) to save
+   e. Move to the next Feature
+3. Review the entire Section 7 after all Features are complete
+   </workflow>
+
+<questions repeat="each_feature">
+1. User Story: "As a [role], I want to [action] so that [benefit]"
+2. What is the user flow?
+3. What are the technical implementation details?
+4. What are the edge cases and error handling strategies?
+5. What are the acceptance criteria?
+</questions>
+
+<example>
+### 7.1 Feature A (F1: Template-based AI Document Writing)
+
+#### 7.1.1 User Story
+
+- When a user starts a conversation, AI guides document writing based on the PRD template
+- Users complete the document step by step following AI guidance
+
+#### 7.1.2 UI Flow
+
+1. Welcome message and getting-started guide on app launch
+2. AI presents questions for the first chapter when user wants to create a new document
+3. After all chapters are complete, guide for final document save
+
+#### 7.1.3 Technical Description
+
+- **System Prompt**: Provide template integrated into LLM
+- **LLM Integration**: Use Amazon Bedrock's Claude Sonnet 4.6 model
+  </example>
+
+<important>
+- Must map 1:1 with requirement IDs from Section 6
+- Each 7.x subsection must be individually confirmed and saved
+- Do not write multiple Features at once
+</important>
+<completion>Save after each Feature is completed; ensure 7.1, 7.2... correspond to all F1, F2...</completion>
+</section_guide>

package/dist/guides/08.md

@@ -0,0 +1,29 @@
+<section_guide number="8" title="MVP Metrics" references="2,6">
+<purpose>Define data collection/analysis methods and success thresholds</purpose>
+
+<required_review>
+MUST review referenced sections before writing:
+
+- Section 2 (MVP Goals): Call read_prd_section(2) for KPIs to measure
+- Section 6.2 (Non-Functional Requirements): Call read_prd_section(6) for NFRs to validate
+  </required_review>
+
+<questions>
+1. How will you measure each goal from Section 2?
+2. What are the data collection methods?
+3. What are the criteria for determining success or failure?
+</questions>
+
+<example>
+### 8.1 Data to Collect
+- **Average document writing time**: Time from when a user starts writing to completion
+- **User command usage frequency**: Number of times commands like `/search` are used
+
+### 8.2 Data Collection Methods
+
+- Record key events in application logs
+- Monitor user interactions via Chainlit's built-in event handlers
+  </example>
+
+<completion>Define measurement methods and thresholds for each KPI</completion>
+</section_guide>

package/dist/guides/09.md

@@ -0,0 +1,17 @@
+<section_guide number="9" title="Out of Scope">
+<purpose>Roadmap for features and technical debt to address in future iterations</purpose>
+
+<questions>
+1. What features are excluded from the MVP?
+2. What are the future improvement plans?
+3. What is the known technical debt?
+</questions>
+
+<example>
+- **S3 File Storage**: Currently supports local file system only; S3 storage integration planned for the future
+- **Document Template Selection**: Expand to allow selection from multiple templates
+- **ECS Deployment**: Currently runs locally; scalable deployment via AWS ECS planned
+</example>
+
+<completion>Organize excluded items and future roadmap</completion>
+</section_guide>

package/dist/index.d.ts

@@ -0,0 +1,2 @@
+#!/usr/bin/env node
+export {};

package/dist/index.js

@@ -0,0 +1,103 @@
+#!/usr/bin/env node
+import { McpServer } from "@modelcontextprotocol/sdk/server/mcp.js";
+import { StdioServerTransport } from "@modelcontextprotocol/sdk/server/stdio.js";
+import { z } from "zod";
+import { TemplateService } from "./tools/templates/service.js";
+import { TemplateController } from "./tools/templates/controller.js";
+import { DocumentService } from "./tools/documents/service.js";
+import { DocumentController } from "./tools/documents/controller.js";
+const server = new McpServer({ name: "prd-writer", version: "0.1.0" }, {
+    instructions: `You are an intelligent product owner helping users create PRD documents.
+
+<WORKFLOW>
+1. init_prd_document() or load_prd_document()
+2. get_prd_overview() - MUST call first to get conversation guide
+3. For each section 1-9:
+   a. get_prd_section_guide(N)
+   b. get_prd_section(N)
+   c. Follow conversation guide from overview
+   d. save_prd_section(N, content) after user confirmation
+5. export_prd_markdown() for final output
+</WORKFLOW>
+
+<RULES>
+- MUST call get_prd_overview() first to get detailed conversation guide
+- NEVER generate multiple sections at once
+- NEVER proceed without user confirmation
+</RULES>`,
+});
+const tc = new TemplateController(new TemplateService());
+const dc = new DocumentController(new DocumentService());
+// Template tools
+server.tool("get_prd_overview", "Get the PRD template overview with all section descriptions. IMPORTANT: After calling this, you MUST call get_prd_section_guide(1) to start the interactive Q&A process.", {}, () => ({
+    content: [{ type: "text", text: tc.getPrdOverview() }],
+}));
+server.tool("list_prd_sections", "List all available PRD template sections.", {}, () => ({
+    content: [{ type: "text", text: JSON.stringify(tc.listPrdSections()) }],
+}));
+server.tool("get_prd_section", "Get a specific PRD template section by number.", {
+    section: z.number().min(1).max(9).describe("Section number (1-9)"),
+    include_examples: z.boolean().default(false).describe("Include example content"),
+}, ({ section, include_examples }) => ({
+    content: [{ type: "text", text: tc.getPrdSection(section, include_examples) }],
+}));
+server.tool("get_prd_full_template", "Get the complete PRD template with all sections combined.", { include_examples: z.boolean().default(false).describe("Include example content") }, ({ include_examples }) => ({
+    content: [{ type: "text", text: tc.getPrdFullTemplate(include_examples) }],
+}));
+server.tool("get_prd_section_guide", "Get conversation guide for writing a specific PRD section. Use this before starting each section.", { section: z.number().min(1).max(9).describe("Section number (1-9)") }, ({ section }) => ({
+    content: [{ type: "text", text: tc.getPrdSectionGuide(section) }],
+}));
+// Document tools
+server.tool("init_prd_document", "Initialize a new PRD document file.", {
+    project_name: z.string().describe("Name of the project"),
+    output_path: z
+        .string()
+        .describe("File path for the document (e.g., ~/Documents/my-project.prd.xml)"),
+}, ({ project_name, output_path }) => ({
+    content: [{ type: "text", text: dc.initPrdDocument(project_name, output_path) }],
+}));
+server.tool("load_prd_document", `Load an existing PRD document to resume editing.
+⚠️ CRITICAL: After loading, you MUST follow the conversation guide:
+1. Call get_prd_section_guide(N) for the section you want to work on
+2. Ask 1-2 focused questions at a time - DO NOT auto-generate content
+3. Wait for user response before proceeding
+4. Get explicit confirmation before saving each section`, { doc_path: z.string().describe("Path to the .prd.xml file") }, ({ doc_path }) => ({
+    content: [{ type: "text", text: dc.loadPrdDocument(doc_path) }],
+}));
+server.tool("save_prd_section", `Save content to a subsection in the PRD document.
+⚠️ BEFORE CALLING THIS TOOL:
+1. 작성 완료된 내용을 사용자에게 먼저 출력하세요
+2. "수정할 내용이 있으신가요?" 라고 확인을 요청하세요
+3. 사용자가 확인한 후에만 이 도구를 호출하세요`, {
+    section: z.number().min(1).max(9).describe("Section number (1-9)"),
+    subsection_id: z.string().describe('Subsection ID (e.g., "1" for X.1, "1.2" for X.1.2)'),
+    title: z.string().describe("Title of the subsection"),
+    content: z.string().describe("Content for the subsection (markdown)"),
+}, ({ section, subsection_id, title, content }) => ({
+    content: [{ type: "text", text: dc.savePrdSection(section, subsection_id, title, content) }],
+}));
+server.tool("read_prd_section", "Read the current content of a section or subsection.", {
+    section: z.number().min(1).max(9).describe("Section number (1-9)"),
+    subsection_id: z
+        .string()
+        .optional()
+        .describe('Subsection ID (e.g., "1" for X.1). If omitted, returns entire section.'),
+}, ({ section, subsection_id }) => ({
+    content: [{ type: "text", text: dc.readPrdSection(section, subsection_id) }],
+}));
+server.tool("get_prd_document_status", "Get the status of all sections in the current document.", {}, () => ({
+    content: [{ type: "text", text: dc.getPrdDocumentStatus() }],
+}));
+server.tool("export_prd_markdown", "Export the PRD document as clean markdown.", {
+    output_path: z
+        .string()
+        .optional()
+        .describe("Optional output file path. If not provided, returns the content."),
+}, ({ output_path }) => ({
+    content: [{ type: "text", text: dc.exportPrdMarkdown(output_path) }],
+}));
+async function main() {
+    const transport = new StdioServerTransport();
+    await server.connect(transport);
+}
+main().catch(console.error);