Production-ready AI workflow orchestration platform with database-driven workflows, personal libraries, and team collaboration.

Agents Playbook empowers AI to handle complex tasks and navigate large codebases with precision. Build custom workflows and mini-prompts for your team's specific needs, maintain personal libraries of reusable patterns, and reduce AI hallucinations through structured, validated processes.

• Database-First Architecture - All workflows and mini-prompts stored in PostgreSQL, no file system access at runtime
• Personal Libraries - Build and maintain your own library of reusable workflows and mini-prompts
• Team Workflows - Create, share, and collaborate on custom workflows tailored to your team's needs
• Visual Workflow Builder - Drag-and-drop constructor for creating sophisticated workflows without YAML
• AI-Powered Discovery - Semantic search using OpenAI embeddings finds the perfect workflow for any task
• MCP Server Integration - Native Model Context Protocol support for Claude Code, Cursor, and other AI assistants
• Authentication & API Tokens - Secure user accounts with NextAuth v5 and API token authentication
• Smart Execution Plans - Automatic injection of Memory Board and Multi-Agent Chat prompts
• Privacy Controls - Public/private workflows with shareable time-limited links

• Node.js 18.0.0 or higher
• npm 8.0.0 or higher
• PostgreSQL database (Neon recommended)
• OpenAI API Key (for semantic search)

# Clone the repository
git clone https://github.com/anthropics/agents-playbook
cd agents-playbook

# Install dependencies
npm install

Create a .env.local file:

# Database (Neon PostgreSQL)
DATABASE_URL=postgresql://...         # Pooled connection
DIRECT_URL=postgresql://...           # Direct connection for migrations

# OpenAI (for semantic search)
OPENAI_API_KEY=sk-...

# NextAuth
NEXTAUTH_SECRET=...                   # Generate with: openssl rand -base64 32
NEXTAUTH_URL=http://localhost:3012    # Your app URL

# Run migrations
npm run db:migrate:dev

# Seed system workflows and mini-prompts
npm run db:seed:system

# Generate embeddings for semantic search
npm run build:embeddings

# Start development server
npm run dev

# Server available at:
# - Web UI: http://localhost:3012
# - MCP Endpoint: http://localhost:3012/api/v1/mcp

Everything is stored in PostgreSQL - no runtime file system access:

• System Workflows & Mini-Prompts: Pre-built content seeded once, available to all users
• User Content: Created through the visual workflow constructor, stored with privacy controls
• Embeddings: OpenAI vectors stored in database for fast semantic search
• Execution Plans: Built dynamically with automatic prompt injection

agents-playbook/
├── prisma/
│   ├── schema.prisma         # Database schema (Workflows, MiniPrompts, Users, etc.)
│   └── migrations/           # Database migrations
├── src/
│   ├── app/                  # Next.js 15 app router
│   │   ├── api/              # API routes (MCP, admin, auth)
│   │   └── dashboard/        # Dashboard pages
│   ├── features/             # Feature-specific components
│   │   ├── workflow-constructor/  # Drag-and-drop workflow builder
│   │   ├── api-tokens/       # API token management
│   │   └── sharing/          # Workflow sharing
│   ├── lib/
│   │   ├── workflows/        # UnifiedWorkflowService, semantic search
│   │   ├── mcp-tools-db/     # Database-backed MCP tools
│   │   ├── auth/             # NextAuth configuration
│   │   └── db/               # Prisma client
│   └── shared/               # Shared UI components
├── scripts/
│   ├── seed-system-content.ts    # Seed system workflows/mini-prompts
│   └── build-embeddings.ts       # Generate OpenAI embeddings
└── tests/                    # Integration & E2E tests

# Run all tests
npm test

# Run integration tests
npm run test:integration

# Run E2E tests
npm run test:e2e

# Run E2E tests in UI mode
npm run test:e2e:ui

1. Discover Workflows: Browse public workflows at /dashboard/discover (no login required)
2. Search with AI: Use semantic search to find workflows by describing your task
3. Create Your Own: Build custom workflows with the visual constructor at /dashboard/library/workflows/new
4. Personal Library: Manage your workflows and mini-prompts at /dashboard/library
5. Team Sharing: Share workflows with time-limited links or make them public

The MCP endpoint (/api/v1/mcp) provides these tools:

• get_available_workflows(task_description) - AI-powered semantic search
• select_workflow(workflow_id) - Get complete workflow with execution plan
• get_next_step(workflow_id, current_step, available_context) - Step-by-step navigation
• get_prompts(search) - Search mini-prompts library
• get_selected_prompt(prompt_id) - Get specific mini-prompt details

Authentication:

• Unauthenticated: Returns system workflows/prompts only
• Authenticated: Returns system + your personal content

💡 Tip: Ask your AI assistant to "use agents-playbook to select workflow" for structured development processes.

1. Create account at agents-playbook.com
2. Navigate to Settings → API Tokens
3. Create new token and copy it

Add to Cursor MCP settings (File → Preferences → Cursor Settings → Features → MCP):

{
  "mcpServers": {
    "agents-playbook": {
      "url": "https://agents-playbook.com/api/v1/mcp",
      "description": "AI Agent Workflow Engine",
      "headers": {
        "Authorization": "Bearer your-api-token"
      }
    }
  }
}

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "agents-playbook": {
      "url": "https://agents-playbook.com/api/v1/mcp",
      "headers": {
        "Authorization": "Bearer your-api-token"
      }
    }
  }
}

Test the MCP server locally:

# Start dev server
npm run dev

# Test with MCP Inspector (no auth)
DANGEROUSLY_OMIT_AUTH=true npx @modelcontextprotocol/inspector@latest http://localhost:3012/api/v1/mcp

Create sophisticated workflows through drag-and-drop interface:

• Add stages (Analysis, Implementation, Testing, etc.)
• Drag mini-prompts from your library into stages
• Configure automatic prompts (Memory Board, Multi-Agent Chat)
• Set privacy controls (PUBLIC/PRIVATE)
• Share with time-limited links

Build your knowledge base:

• Workflows: Multi-step processes for complex tasks
• Mini-Prompts: Reusable prompt templates
• Privacy Controls: Keep private or share publicly
• Tagging: Organize with custom tags
• Search: AI-powered semantic search across your library

Workflows can include two automatic prompts:

• Memory Board (📋): Injected at end of each stage for context handoff
• Multi-Agent Chat (🤖): Coordination prompts after each mini-prompt
• Admins can edit at /dashboard/admin/system-prompts

Browse public workflows without login:

• Semantic search: Describe your task in natural language
• Filter by complexity, tags, ratings
• Preview execution plans before using
• Fork workflows to your library

npm i -g vercel
vercel --prod

Set environment variables in Vercel dashboard:

• DATABASE_URL - Pooled PostgreSQL connection
• DIRECT_URL - Direct PostgreSQL connection
• OPENAI_API_KEY - For semantic search
• NEXTAUTH_SECRET - Session encryption key
• NEXTAUTH_URL - Production URL

No workflows found:

• Run npm run db:seed:system to populate system workflows
• Run npm run build:embeddings to generate embeddings
• Check database: npm run db:studio

Database connection errors:

• Verify DATABASE_URL includes ?sslmode=require
• Use DIRECT_URL for migrations
• Regenerate Prisma client: npm run db:generate

MCP server not responding:

• Ensure dev server running: npm run dev
• Check endpoint: http://localhost:3012/api/v1/mcp
• Verify API token in headers

Build failures:

• Clear Next.js cache: rm -rf .next
• Regenerate Prisma client: npm run db:generate
• Check TypeScript: npm run lint

MIT License - see LICENSE file for details.