{
  "markdown": "# AI Agent Tutorials & Implementations\n\nA comprehensive collection of production-ready AI agent implementations showcasing different frameworks, protocols, and integration patterns. This repository demonstrates various approaches to building intelligent agents with Model Context Protocol (MCP), multi-agent systems, and real-world integrations.\n\n## Repository Overview\n\nThis repository contains multiple agent implementations, each demonstrating different architectural patterns and use cases:\n\n| Project | Framework | Key Features | Use Case |\n|---------|-----------|--------------|----------|\n| [agent2agent](#agent2agent) | LangGraph + A2A Protocol | Remote agent communication, Slack integration | Investment research |\n| [mcp-financial](#mcp-financial) | FastMCP + FastAPI | ASGI integration, CLI client | Financial data analysis |\n| [bright-mcp-server-overview](#bright-mcp-server-overview) | Dual: LangGraph + ADK | Memory persistence, extended timeouts | Web scraping & research |\n| [fpl-deepagent](#fpl-deepagent) | FastMCP + React UI | Streamable HTTP, ChatGPT integration | Fantasy Premier League |\n| [task-manager-app](#task-manager-app) | FastMCP + React UI + Supabase | OAuth (Auth0), per-user DB state, Slack notifications | Task management in ChatGPT |\n| [notion-mcp-agent](#notion-mcp-agent) | LangGraph + MCP | Notion integration, database management | Knowledge management |\n| [claude-advanced-tool-use](#claude-advanced-tool-use) | Claude API + FastMCP | PTC, Tool Search, MCP integration | Token-efficient AI agents |\n| [claude-skills](#claude-skills) | Claude Skills API | Document generation, custom skills | PowerPoint, Excel, Word creation |\n| [openai-chatkit-starter-app](#openai-chatkit-starter-app) | Next.js + ChatKit | Agent Builder integration, web component | ChatKit UI development |\n| [mastra-overview](#mastra-overview) | Mastra framework | Multi-LLM orchestration | Framework exploration |\n| [smithery-example](#smithery-example) | Smithery + FastMCP | MCP playground, development tools | MCP development |\n| [mcp-apps](#mcp-apps) | MCP Apps (OpenAI Apps SDK) | Example MCP Apps (weather + stock analysis) | MCP Apps reference implementations |\n\n## Project Descriptions\n\n### agent2agent/\n**Investment Research Analyst Agent**\n\nA production-ready investment research agent implementing Google's Agent-to-Agent (A2A) protocol for remote agent communication.\n\n**Key Features:**\n- **Framework**: LangGraph with LangChain\n- **Protocol**: Agent-to-Agent (A2A) for remote communication\n- **Integration**: Slack with Block Kit UI and metadata modals\n- **Architecture**: FastAPI server exposing both A2A endpoints and Slack events\n- **Memory**: Persistent conversation state management\n- **Deployment**: Docker ready with Render.com configuration\n\n**Technical Stack:**\n- LangGraph for agent orchestration\n- FastAPI for A2A protocol implementation\n- Slack Block Kit for interactive UI\n- LangSmith for observability (optional)\n- Docker for containerized deployment\n\n**Use Cases:**\n- Stock summaries and analysis\n- SEC filings research\n- Analyst recommendations\n- Financial data aggregation\n- Investment research workflows\n\n### mcp-financial/\n**Investment Analyst MCP Agent**\n\nA financial data agent powered by FastMCP with ASGI integration, providing both CLI and Slack interfaces.\n\n**Key Features:**\n- **Framework**: FastMCP with FastAPI ASGI integration\n- **Interfaces**: CLI client and Slack bot\n- **Architecture**: MCP server exposed via FastAPI endpoints\n- **Integration**: Direct Slack event handling\n- **Deployment**: Production-ready with health checks\n\n**Technical Stack:**\n- FastMCP for Model Context Protocol implementation\n- FastAPI for ASGI integration\n- Uvicorn for server runtime\n- Slack API for bot functionality\n- MCP Inspector for debugging\n\n**Use Cases:**\n- Financial data analysis\n- Stock price monitoring\n- Earnings analysis\n- Market research\n- Investment insights\n\n### bright-mcp-server-overview/\n**Bright Data MCP Research Agent**\n\nA comprehensive research agent powered by Bright Data's web scraping infrastructure, featuring dual AI agent implementations.\n\n**Key Features:**\n- **Dual Framework**: LangGraph (with memory) + Google ADK (with extended timeouts)\n- **Integration**: Bright Data MCP server for web scraping\n- **Slack Interface**: Interactive agent selection via dropdown\n- **Memory**: Persistent conversation memory (LangGraph)\n- **Timeouts**: Extended timeout handling (ADK) for long operations\n- **Specialization**: SEO research, e-commerce intelligence, market analysis\n\n**Technical Stack:**\n- **LangGraph Agent**: OpenAI GPT with MemorySaver checkpointer\n- **ADK Agent**: Google Gemini 2.0 Flash with custom timeout patches\n- **MCP Integration**: Bright Data MCP server for data collection\n- **Slack Integration**: Bot with agent selection and interactive UI\n\n**Agent Comparison:**\n| Feature | LangGraph Agent | ADK Agent |\n|---------|----------------|-----------|\n| Memory | Persistent (checkpointer) | Context-aware (5 messages) |\n| Timeout | Standard (5s) | Extended (60s) |\n| Model | OpenAI GPT | Gemini 2.0 Flash |\n| Best For | Interactive conversations | Long-running operations |\n\n**Use Cases:**\n- SEO keyword research and SERP analysis\n- E-commerce product monitoring and price tracking\n- Competitor analysis and market intelligence\n- Web scraping and data collection\n- Business intelligence and insights\n\n### fpl-deepagent/\n**Fantasy Premier League MCP Assistant**\n\nA comprehensive Fantasy Premier League assistant that integrates with ChatGPT through the Model Context Protocol (MCP), featuring beautiful React UI components and real-time FPL data.\n\n**Key Features:**\n- **Framework**: FastMCP with Streamable HTTP transport\n- **UI Integration**: React 18 + TypeScript components for ChatGPT\n- **Real-time Data**: Live FPL API integration with caching and error handling\n- **Design Compliance**: Follows OpenAI Apps SDK design guidelines exactly\n- **Interactive Tools**: Player search, detailed stats, and side-by-side comparison\n\n**Technical Stack:**\n- FastMCP for MCP server implementation\n- React 18 + TypeScript for UI components\n- OpenAI Apps SDK integration with `window.openai` API\n- esbuild for fast, modern bundling\n- Streamable HTTP for bidirectional communication\n\n**UI Components:**\n- **PlayerListComponent**: Interactive player grid with favorites\n- **PlayerDetailComponent**: Detailed player stats and upcoming fixtures\n- **PlayerComparisonComponent**: Side-by-side comparison with highlighted stats\n\n**Use Cases:**\n- Player search and discovery\n- Detailed player statistics and form analysis\n- Player comparison for team selection\n- FPL team optimization\n- Real-time price and form tracking\n\n### task-manager-app/\n**Task Manager ChatGPT App (Apps SDK + MCP + Supabase + OAuth)**\n\nA production-ready tutorial showing how to build a ChatGPT App with:\n- **FastMCP (Streamable HTTP)** as the MCP server\n- **React widgets** rendered inside ChatGPT\n- **Supabase (Postgres)** as authoritative state for tasks/notifications\n- **OAuth (Auth0)** for multi-user authentication (MCP OAuth)\n- Optional **Slack notifications** (send now + schedule)\n\nStart here:\n- `task-manager-app/README.md`\n\n### notion-mcp-agent/\n**Notion Knowledge Management Agent**\n\nA sophisticated agent that integrates with Notion through MCP, providing intelligent database management and knowledge organization capabilities.\n\n**Key Features:**\n- **Framework**: LangGraph with MCP integration\n- **Integration**: Notion API for database operations\n- **Slack Interface**: Interactive knowledge management\n- **Context Management**: Intelligent data aggregation\n- **Database Operations**: Create, read, update, and organize Notion databases\n\n**Technical Stack:**\n- LangGraph for agent orchestration\n- Notion MCP server for database operations\n- Slack API for user interaction\n- Context aggregation for intelligent responses\n\n**Use Cases:**\n- Knowledge base management\n- Database organization and maintenance\n- Content aggregation and structuring\n- Team collaboration workflows\n- Information retrieval and organization\n\n### claude-advanced-tool-use/\n**Claude Advanced Tool Use Tutorial**\n\nA comprehensive tutorial demonstrating Anthropic's Advanced Tool Use features: Programmatic Tool Calling (PTC) and Tool Search. These features enable AI agents to scale to thousands of tools while dramatically reducing token usage.\n\n**Key Features:**\n- **Programmatic Tool Calling (PTC)**: Claude writes Python code that orchestrates tool calls in a sandbox\n- **Tool Search**: Dynamic tool discovery with `defer_loading` for efficient context usage\n- **MCP Integration**: Tool Search combined with MCP servers via `mcp_toolset`\n- **Real-World Examples**: Financial data tools using yfinance\n- **Token Savings**: Up to 98% reduction in token usage for complex tasks\n\n**Technical Stack:**\n- Anthropic Claude API (Sonnet 4.5)\n- Beta headers: `advanced-tool-use-2025-11-20`\n- FastMCP for MCP server implementation\n- Python + yfinance for financial data\n- ngrok for MCP server tunneling\n\n**Examples:**\n- `01_ptc_token_savings.py` - Programmatic Tool Calling with token comparison\n- `02_tool_search.py` - Tool Search with 10 deferred financial tools\n- `03_mcp_tool_search.py` - MCP + Tool Search via ngrok tunnel\n- `mcp_server.py` - FastMCP server exposing financial tools\n\n**Key Concepts:**\n| Feature | Description | Token Savings |\n|---------|-------------|---------------|\n| Programmatic Tool Calling | Tool results stay in sandbox, only `print()` output enters context | 37% |\n| Tool Search | Only load tool definitions when discovered | 85% |\n| Combined | PTC + Tool Search together | Up to 98% |\n\n**Use Cases:**\n- Building AI agents with many tools (100+)\n- Reducing context window bloat from tool definitions\n- Processing large datasets without context overflow\n- MCP server integration with dynamic tool discovery\n- Token-efficient financial analysis agents\n\n### claude-skills/\n**Claude Skills API Implementation**\n\nA comprehensive implementation of Claude's Skills API for automated document generation and custom skill creation.\n\n**Key Features:**\n- **Framework**: Claude Skills API with streaming support\n- **Document Generation**: PowerPoint, Excel, Word, and PDF creation\n- **Custom Skills**: Upload and manage custom skills (8MB limit)\n- **File Management**: List, download, and delete generated files\n- **Multi-Skill Workflows**: Combine multiple skills in single requests\n\n**Technical Stack:**\n- Claude Skills API with beta features\n- Code execution environment (2025-08-25)\n- Files API (2025-04-14)\n- Streaming responses for real-time progress\n- Python SDK with uv package manager\n\n**Utilities:**\n- `list-skills.py` - List all available skills\n- `create-skill.py` - Upload custom skills from directories\n- `use-skill.py` - Generate documents with single skills\n- `multi-skill-demo.py` - Complex workflows with multiple skills\n- `list-files.py` / `download-file.py` / `delete-file.py` - File management\n\n**Use Cases:**\n- Automated PowerPoint presentation generation\n- Excel spreadsheet creation and data analysis\n- Word document generation\n- PDF report creation\n- Custom skill development and deployment\n- Multi-format document workflows\n\n### openai-chatkit-starter-app/\n**ChatKit Web Component Starter**\n\nA minimal Next.js starter template for building ChatKit applications with OpenAI's Agent Builder workflows.\n\n**Key Features:**\n- **Framework**: Next.js with ChatKit web component\n- **Integration**: OpenAI Agent Builder workflows\n- **Customization**: Configurable themes, prompts, and UI\n- **Session Management**: Ready-to-use session endpoint\n- **Deployment**: Domain allowlist verification support\n\n**Technical Stack:**\n- Next.js for application framework\n- OpenAI ChatKit web component (`<openai-chatkit>`)\n- OpenAI API integration\n- TypeScript for type safety\n- Configurable theming system\n\n**Key Components:**\n- Session creation endpoint (`/api/create-session`)\n- ChatKit panel with event handlers\n- Theme and color scheme controls\n- Starter prompts configuration\n- Error overlay for debugging\n\n**Use Cases:**\n- ChatKit application prototyping\n- Agent Builder workflow integration\n- Custom ChatKit UI development\n- OpenAI workflow testing\n- Production ChatKit deployments\n\n### mastra-overview/\n**Mastra Framework Exploration**\n\nAn exploration of the Mastra framework for multi-LLM orchestration and agent management.\n\n**Key Features:**\n- **Framework**: Mastra for multi-LLM orchestration\n- **Multi-LLM**: Support for multiple language models\n- **Orchestration**: Intelligent model selection and routing\n- **Polyfills**: Crypto polyfills for browser compatibility\n\n**Technical Stack:**\n- Mastra framework\n- Multi-LLM integration\n- Browser compatibility polyfills\n- TypeScript configuration\n\n**Use Cases:**\n- Multi-LLM agent systems\n- Model orchestration and routing\n- Framework exploration and evaluation\n- LLM comparison and benchmarking\n\n### smithery-example/\n**MCP Development Playground**\n\nA comprehensive development environment for MCP (Model Context Protocol) with FastMCP integration and testing tools.\n\n**Key Features:**\n- **Framework**: Smithery + FastMCP\n- **Development Tools**: MCP playground and testing environment\n- **Financial Integration**: Example financial server implementation\n- **Testing**: Comprehensive test suite and examples\n- **Documentation**: Development guides and examples\n\n**Technical Stack:**\n- Smithery for MCP development\n- FastMCP for server implementation\n- Testing frameworks for validation\n- Development tooling and playgrounds\n\n**Use Cases:**\n- MCP server development\n- Protocol testing and validation\n- Financial data integration examples\n- Development environment setup\n- MCP learning and exploration\n\n### mcp-apps/\n**MCP Apps Examples (Weather + Stock Analysis)**\n\nTwo minimal example MCP Apps showing how to build UI + server experiences using the MCP Apps extensions.\n\n**Key Features:**\n- **Weather App**: UI + MCP server example with a simple weather workflow\n- **Stock Analysis App**: UI + MCP server example for market/stock analysis\n- **Apps SDK**: Designed to follow MCP Apps extension patterns\n- **Docs Reference**: See the MCP Apps docs for the full guide\n\n**Use Cases:**\n- Learning MCP Apps fundamentals\n- Building UI-backed MCP Apps\n- Reference implementations for new MCP App projects\n\n## Getting Started\n\nEach project includes comprehensive setup instructions in its respective README file. General prerequisites include:\n\n### Common Requirements\n- Python 3.9+ (some projects require newer; see each project README)\n- Valid API keys for respective services\n- Slack workspace access (for Slack integrations)\n- Environment variable configuration\n\n### Quick Start Pattern\n```bash\n# 1. Navigate to desired project\ncd [project-name]/\n\n# 2. Install dependencies\n# Most Python projects here use uv:\nuv sync\n# Some projects use pip/requirements.txt:\n# pip install -r requirements.txt\n\n# 3. Configure environment\ncp .env.example .env\n# Edit .env with your API keys\n\n# 4. Run the agent\n# (varies by project - see individual READMEs)\n```\n\n## Architecture Patterns\n\n### Model Context Protocol (MCP)\nMultiple projects demonstrate different MCP implementation patterns:\n- **FastMCP ASGI**: Direct FastAPI integration (mcp-financial, smithery-example)\n- **FastMCP Streamable HTTP**: Modern bidirectional communication (fpl-deepagent)\n- **Bright Data MCP**: External MCP server communication\n- **Notion MCP**: Database and knowledge management integration\n\n### Agent Communication\n- **A2A Protocol**: Remote agent-to-agent communication (agent2agent)\n- **State Management**: Persistent conversation memory (bright-mcp-server-overview)\n\n### UI Integration Patterns\n- **React + ChatGPT**: OpenAI Apps SDK integration (fpl-deepagent)\n- **Next.js + ChatKit**: Agent Builder workflow integration (openai-chatkit-starter-app)\n- **Slack Bots**: Event-driven chat interfaces (multiple projects)\n- **CLI Clients**: Command-line agent interaction\n\n### Document Generation\n- **Claude Skills API**: Automated document creation with streaming (claude-skills)\n- **Multi-Format Support**: PowerPoint, Excel, Word, PDF generation\n- **Custom Skills**: Uploadable skill packages for specialized tasks\n\n### Development & Testing\n- **MCP Playground**: Development and testing environment (smithery-example)\n- **Multi-LLM Orchestration**: Framework exploration (mastra-overview)\n- **Agent Builder**: OpenAI workflow development (openai-chatkit-starter-app)\n\n### Integration Patterns\n- **Container Deployment**: Docker and cloud-ready\n- **API Integration**: RESTful agent endpoints\n- **Database Integration**: Knowledge management systems\n- **Real-time Data**: Live API integration with caching\n\n## Contributing\n\nEach project welcomes contributions. Please:\n\n1. Fork the repository\n2. Create a feature branch\n3. Follow the project's coding standards\n4. Include tests where applicable\n5. Submit a Pull Request\n\n## License\n\nMIT License - see individual project LICENSE files for details.\n\n## Support & Resources\n\n### Documentation Links\n- [Model Context Protocol](https://modelcontextprotocol.io/)\n- [LangGraph Documentation](https://langchain-ai.github.io/langgraph/)\n- [OpenAI Agent SDK](https://github.com/openai/agent-sdk)\n- [OpenAI Apps SDK](https://developers.openai.com/apps-sdk/)\n- [OpenAI ChatKit](http://openai.github.io/chatkit-js/)\n- [OpenAI Agent Builder](https://platform.openai.com/agent-builder)\n- [Claude Skills API](https://docs.claude.com/en/api/skills-guide)\n- [Claude Programmatic Tool Calling](https://platform.claude.com/docs/en/agents-and-tools/tool-use/programmatic-tool-calling)\n- [Claude Tool Search](https://platform.claude.com/docs/en/agents-and-tools/tool-use/tool-search-tool)\n- [Anthropic Blog - Advanced Tool Use](https://www.anthropic.com/engineering/advanced-tool-use)\n- [Anthropic Blog - Code Execution](https://www.anthropic.com/engineering/code-execution-with-mcp)\n- [Anthropic Console](https://console.anthropic.com/)\n- [Google ADK](https://developers.google.com/ai/adk)\n- [FastMCP](https://github.com/pydantic/fastmcp)\n- [Mastra Framework](https://mastra.ai/)\n- [Smithery](https://smithery.ai/)\n- [Slack API](https://api.slack.com/)\n\n### Platform-Specific Support\n- **Bright Data**: [brightdata.com/support](https://brightdata.com/support)\n- **Notion**: [developers.notion.com](https://developers.notion.com/)\n- **Fantasy Premier League**: [fpl.readthedocs.io](https://fpl.readthedocs.io/en/latest/)\n- **Slack**: [api.slack.com/support](https://api.slack.com/support)\n\n---\n\n**Built with ❤️ demonstrating the future of AI agent development**\n",
  "bytes": 18552,
  "sha": "b2354219e4105a2086a6ec232da1f64905123a85d6cef93bd6864457a9c5284d",
  "repo_slug": "hollaugo/tutorials",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/mcp_ai_smithery_hollaugo_financial_research__2254858d/readme"
}