@keeborg/mcp

MCP (Model Context Protocol) server for Keeborg - Generate complete technical specifications for AI coding agents in under 60 seconds.

What is Keeborg?

Keeborg generates an 8-document specification suite from a simple app idea:

• PRD - Product Requirements Document with features and user stories
• Architecture - System design, tech stack, and component diagrams
• API Spec - Complete endpoint definitions with request/response schemas
• DB Schema - Database tables, relationships, and indexes
• UX Spec - User flows, states, and navigation maps
• UI Spec - Screen layouts, components, and design tokens
• Implementation Guide - Security, deployment, and testing requirements
• Agent Workflow - Phased development plan optimized for AI agents

Quick Start

1. Install

npm install -g @keeborg/mcp

2. Get Your API Key

1. Sign up at keeborg.com
2. Go to Settings > API Keys
3. Create a new key (starts with kb_live_)

3. Configure Your AI Agent

Claude Code

Add to ~/.claude/settings.json:

{
  "mcpServers": {
    "keeborg": {
      "command": "keeborg-mcp",
      "env": {
        "KEEBORG_API_KEY": "kb_live_your_key_here"
      }
    }
  }
}

Cursor

Add to your Cursor MCP settings:

{
  "keeborg": {
    "command": "keeborg-mcp",
    "env": {
      "KEEBORG_API_KEY": "kb_live_your_key_here"
    }
  }
}

Available Tools

| Tool | Description | |------|-------------| | create_project | Create a new project and generate specs | | get_project_status | Check generation progress | | get_document | Retrieve a specific spec document | | list_projects | List all your projects | | get_prompts | Get AI-optimized prompts from specs | | regenerate_document | Refresh a single document |

create_project

Create a new project and generate technical specifications.

Arguments:
  name (required)        Short project name
  description (required) Detailed description of what to build
  tech_stack (optional)  Tech preferences object:
    - frontend_framework: "Next.js", "React", "Vue", etc.
    - css_styling: "Tailwind CSS", "CSS Modules", etc.
    - backend_framework: "FastAPI", "Express", "Django", etc.
    - database: "PostgreSQL", "MongoDB", "Supabase", etc.
    - auth_provider: "Supabase Auth", "Auth0", "Clerk", etc.
    - deployment: "Vercel", "AWS", "Coolify", etc.

get_project_status

Check generation progress. Poll until status is COMPLETED.

Arguments:
  project_id (required)  Project ID from create_project

get_document

Retrieve a specific specification document.

Arguments:
  project_id (required)  Project ID
  doc_type (required)    One of:
    - PRD
    - ARCHITECTURE
    - API_SPEC
    - DB_SCHEMA
    - UX_SPEC
    - UI_SPEC
    - IMPLEMENTATION_GUIDE
    - AGENT_WORKFLOW

list_projects

List all your projects with their status.

get_prompts

Get AI agent prompts generated from the specs.

Arguments:
  project_id (required)  Project ID
  prompt_type (optional) "primary", "backend", "frontend",
                         "fullstack", "devops", or "testing"

regenerate_document

Regenerate a single document using current project context.

Arguments:
  project_id (required)  Project ID
  doc_type (required)    Document type to regenerate

Example Workflow

You: "I want to build a habit tracking app with React Native and Supabase"

Claude: [Uses create_project]
        "Project created! Generating specifications...
         This typically completes in 30-60 seconds."

        [Polls get_project_status]
        "All 8 documents ready!"

        [Uses get_document with DB_SCHEMA]
        "Based on the database schema, I'll create these tables:
         - users, habits, habit_logs, streaks..."

        [Uses get_document with API_SPEC]
        "The API defines these endpoints:
         - POST /habits, GET /habits, POST /habits/:id/log..."

        [Proceeds to implement using the complete specs]

Error Handling

The MCP server handles API errors gracefully:

| Status | Meaning | Action | |--------|---------|--------| | 504 | LLM generation timed out | Auto-retries up to 2 times with backoff | | 429 | Rate limit exceeded | Wait and retry (5 generates/min, 10 regenerates/min) | | 500 | Server error | Reported to agent with error details |

Requirements

• Node.js 18+
• Keeborg Pro subscription (for API access)

Development

# Clone the repo
git clone https://github.com/keeborg/mcp-server.git
cd mcp-server

# Install dependencies
npm install

# Build
npm run build

# Run locally
KEEBORG_API_KEY=kb_live_xxx npm start

Contributing

Contributions are welcome! Please see CONTRIBUTING.md for guidelines.

Links

• Keeborg Website
• API Documentation
• Interactive Demo - See example specs without signing up
• Changelog

License

MIT - see LICENSE for details.