AI Agent Tutorials & Implementations

A 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.

Repository Overview

This repository contains multiple agent implementations, each demonstrating different architectural patterns and use cases:

Project	Framework	Key Features	Use Case
agent2agent	LangGraph + A2A Protocol	Remote agent communication, Slack integration	Investment research
mcp-financial	FastMCP + FastAPI	ASGI integration, CLI client	Financial data analysis
bright-mcp-server-overview	Dual: LangGraph + ADK	Memory persistence, extended timeouts	Web scraping & research
fpl-deepagent	FastMCP + React UI	Streamable HTTP, ChatGPT integration	Fantasy Premier League
task-manager-app	FastMCP + React UI + Supabase	OAuth (Auth0), per-user DB state, Slack notifications	Task management in ChatGPT
notion-mcp-agent	LangGraph + MCP	Notion integration, database management	Knowledge management
claude-advanced-tool-use	Claude API + FastMCP	PTC, Tool Search, MCP integration	Token-efficient AI agents
claude-skills	Claude Skills API	Document generation, custom skills	PowerPoint, Excel, Word creation
openai-chatkit-starter-app	Next.js + ChatKit	Agent Builder integration, web component	ChatKit UI development
mastra-overview	Mastra framework	Multi-LLM orchestration	Framework exploration
smithery-example	Smithery + FastMCP	MCP playground, development tools	MCP development
mcp-apps	MCP Apps (OpenAI Apps SDK)	Example MCP Apps (weather + stock analysis)	MCP Apps reference implementations

Project Descriptions

agent2agent/

Investment Research Analyst Agent

A production-ready investment research agent implementing Google's Agent-to-Agent (A2A) protocol for remote agent communication.

Key Features:

• Framework: LangGraph with LangChain
• Protocol: Agent-to-Agent (A2A) for remote communication
• Integration: Slack with Block Kit UI and metadata modals
• Architecture: FastAPI server exposing both A2A endpoints and Slack events
• Memory: Persistent conversation state management
• Deployment: Docker ready with Render.com configuration

Technical Stack:

• LangGraph for agent orchestration
• FastAPI for A2A protocol implementation
• Slack Block Kit for interactive UI
• LangSmith for observability (optional)
• Docker for containerized deployment

Use Cases:

• Stock summaries and analysis
• SEC filings research
• Analyst recommendations
• Financial data aggregation
• Investment research workflows

mcp-financial/

Investment Analyst MCP Agent

A financial data agent powered by FastMCP with ASGI integration, providing both CLI and Slack interfaces.

Key Features:

• Framework: FastMCP with FastAPI ASGI integration
• Interfaces: CLI client and Slack bot
• Architecture: MCP server exposed via FastAPI endpoints
• Integration: Direct Slack event handling
• Deployment: Production-ready with health checks

Technical Stack:

• FastMCP for Model Context Protocol implementation
• FastAPI for ASGI integration
• Uvicorn for server runtime
• Slack API for bot functionality
• MCP Inspector for debugging

Use Cases:

• Financial data analysis
• Stock price monitoring
• Earnings analysis
• Market research
• Investment insights

bright-mcp-server-overview/

Bright Data MCP Research Agent

A comprehensive research agent powered by Bright Data's web scraping infrastructure, featuring dual AI agent implementations.

Key Features:

• Dual Framework: LangGraph (with memory) + Google ADK (with extended timeouts)
• Integration: Bright Data MCP server for web scraping
• Slack Interface: Interactive agent selection via dropdown
• Memory: Persistent conversation memory (LangGraph)
• Timeouts: Extended timeout handling (ADK) for long operations
• Specialization: SEO research, e-commerce intelligence, market analysis

Technical Stack:

• LangGraph Agent: OpenAI GPT with MemorySaver checkpointer
• ADK Agent: Google Gemini 2.0 Flash with custom timeout patches
• MCP Integration: Bright Data MCP server for data collection
• Slack Integration: Bot with agent selection and interactive UI

Agent Comparison:

Feature	LangGraph Agent	ADK Agent
Memory	Persistent (checkpointer)	Context-aware (5 messages)
Timeout	Standard (5s)	Extended (60s)
Model	OpenAI GPT	Gemini 2.0 Flash
Best For	Interactive conversations	Long-running operations

Use Cases:

• SEO keyword research and SERP analysis
• E-commerce product monitoring and price tracking
• Competitor analysis and market intelligence
• Web scraping and data collection
• Business intelligence and insights

fpl-deepagent/

Fantasy Premier League MCP Assistant

A 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.

Key Features:

• Framework: FastMCP with Streamable HTTP transport
• UI Integration: React 18 + TypeScript components for ChatGPT
• Real-time Data: Live FPL API integration with caching and error handling
• Design Compliance: Follows OpenAI Apps SDK design guidelines exactly
• Interactive Tools: Player search, detailed stats, and side-by-side comparison

Technical Stack:

• FastMCP for MCP server implementation
• React 18 + TypeScript for UI components
• OpenAI Apps SDK integration with window.openai API
• esbuild for fast, modern bundling
• Streamable HTTP for bidirectional communication

UI Components:

• PlayerListComponent: Interactive player grid with favorites
• PlayerDetailComponent: Detailed player stats and upcoming fixtures
• PlayerComparisonComponent: Side-by-side comparison with highlighted stats

Use Cases:

• Player search and discovery
• Detailed player statistics and form analysis
• Player comparison for team selection
• FPL team optimization
• Real-time price and form tracking

task-manager-app/

Task Manager ChatGPT App (Apps SDK + MCP + Supabase + OAuth)

A production-ready tutorial showing how to build a ChatGPT App with:

• FastMCP (Streamable HTTP) as the MCP server
• React widgets rendered inside ChatGPT
• Supabase (Postgres) as authoritative state for tasks/notifications
• OAuth (Auth0) for multi-user authentication (MCP OAuth)
• Optional Slack notifications (send now + schedule)

Start here:

• task-manager-app/README.md

notion-mcp-agent/

Notion Knowledge Management Agent

A sophisticated agent that integrates with Notion through MCP, providing intelligent database management and knowledge organization capabilities.

Key Features:

• Framework: LangGraph with MCP integration
• Integration: Notion API for database operations
• Slack Interface: Interactive knowledge management
• Context Management: Intelligent data aggregation
• Database Operations: Create, read, update, and organize Notion databases

Technical Stack:

• LangGraph for agent orchestration
• Notion MCP server for database operations
• Slack API for user interaction
• Context aggregation for intelligent responses

Use Cases:

• Knowledge base management
• Database organization and maintenance
• Content aggregation and structuring
• Team collaboration workflows
• Information retrieval and organization

claude-advanced-tool-use/

Claude Advanced Tool Use Tutorial

A 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.

Key Features:

• Programmatic Tool Calling (PTC): Claude writes Python code that orchestrates tool calls in a sandbox
• Tool Search: Dynamic tool discovery with defer_loading for efficient context usage
• MCP Integration: Tool Search combined with MCP servers via mcp_toolset
• Real-World Examples: Financial data tools using yfinance
• Token Savings: Up to 98% reduction in token usage for complex tasks

Technical Stack:

• Anthropic Claude API (Sonnet 4.5)
• Beta headers: advanced-tool-use-2025-11-20
• FastMCP for MCP server implementation
• Python + yfinance for financial data
• ngrok for MCP server tunneling

Examples:

• 01_ptc_token_savings.py - Programmatic Tool Calling with token comparison
• 02_tool_search.py - Tool Search with 10 deferred financial tools
• 03_mcp_tool_search.py - MCP + Tool Search via ngrok tunnel
• mcp_server.py - FastMCP server exposing financial tools

Key Concepts:

Feature	Description	Token Savings
Programmatic Tool Calling	Tool results stay in sandbox, only print() output enters context	37%
Tool Search	Only load tool definitions when discovered	85%
Combined	PTC + Tool Search together	Up to 98%

Use Cases:

• Building AI agents with many tools (100+)
• Reducing context window bloat from tool definitions
• Processing large datasets without context overflow
• MCP server integration with dynamic tool discovery
• Token-efficient financial analysis agents

claude-skills/

Claude Skills API Implementation

A comprehensive implementation of Claude's Skills API for automated document generation and custom skill creation.

Key Features:

• Framework: Claude Skills API with streaming support
• Document Generation: PowerPoint, Excel, Word, and PDF creation
• Custom Skills: Upload and manage custom skills (8MB limit)
• File Management: List, download, and delete generated files
• Multi-Skill Workflows: Combine multiple skills in single requests

Technical Stack:

• Claude Skills API with beta features
• Code execution environment (2025-08-25)
• Files API (2025-04-14)
• Streaming responses for real-time progress
• Python SDK with uv package manager

Utilities:

• list-skills.py - List all available skills
• create-skill.py - Upload custom skills from directories
• use-skill.py - Generate documents with single skills
• multi-skill-demo.py - Complex workflows with multiple skills
• list-files.py / download-file.py / delete-file.py - File management

Use Cases:

• Automated PowerPoint presentation generation
• Excel spreadsheet creation and data analysis
• Word document generation
• PDF report creation
• Custom skill development and deployment
• Multi-format document workflows

openai-chatkit-starter-app/

ChatKit Web Component Starter

A minimal Next.js starter template for building ChatKit applications with OpenAI's Agent Builder workflows.

Key Features:

• Framework: Next.js with ChatKit web component
• Integration: OpenAI Agent Builder workflows
• Customization: Configurable themes, prompts, and UI
• Session Management: Ready-to-use session endpoint
• Deployment: Domain allowlist verification support

Technical Stack:

• Next.js for application framework
• OpenAI ChatKit web component (<openai-chatkit>)
• OpenAI API integration
• TypeScript for type safety
• Configurable theming system

Key Components:

• Session creation endpoint (/api/create-session)
• ChatKit panel with event handlers
• Theme and color scheme controls
• Starter prompts configuration
• Error overlay for debugging

Use Cases:

• ChatKit application prototyping
• Agent Builder workflow integration
• Custom ChatKit UI development
• OpenAI workflow testing
• Production ChatKit deployments

mastra-overview/

Mastra Framework Exploration

An exploration of the Mastra framework for multi-LLM orchestration and agent management.

Key Features:

• Framework: Mastra for multi-LLM orchestration
• Multi-LLM: Support for multiple language models
• Orchestration: Intelligent model selection and routing
• Polyfills: Crypto polyfills for browser compatibility

Technical Stack:

• Mastra framework
• Multi-LLM integration
• Browser compatibility polyfills
• TypeScript configuration

Use Cases:

• Multi-LLM agent systems
• Model orchestration and routing
• Framework exploration and evaluation
• LLM comparison and benchmarking

smithery-example/

MCP Development Playground

A comprehensive development environment for MCP (Model Context Protocol) with FastMCP integration and testing tools.

Key Features:

• Framework: Smithery + FastMCP
• Development Tools: MCP playground and testing environment
• Financial Integration: Example financial server implementation
• Testing: Comprehensive test suite and examples
• Documentation: Development guides and examples

Technical Stack:

• Smithery for MCP development
• FastMCP for server implementation
• Testing frameworks for validation
• Development tooling and playgrounds

Use Cases:

• MCP server development
• Protocol testing and validation
• Financial data integration examples
• Development environment setup
• MCP learning and exploration

mcp-apps/

MCP Apps Examples (Weather + Stock Analysis)

Two minimal example MCP Apps showing how to build UI + server experiences using the MCP Apps extensions.

Key Features:

• Weather App: UI + MCP server example with a simple weather workflow
• Stock Analysis App: UI + MCP server example for market/stock analysis
• Apps SDK: Designed to follow MCP Apps extension patterns
• Docs Reference: See the MCP Apps docs for the full guide

Use Cases:

• Learning MCP Apps fundamentals
• Building UI-backed MCP Apps
• Reference implementations for new MCP App projects

Getting Started

Each project includes comprehensive setup instructions in its respective README file. General prerequisites include:

Common Requirements

• Python 3.9+ (some projects require newer; see each project README)
• Valid API keys for respective services
• Slack workspace access (for Slack integrations)
• Environment variable configuration

Quick Start Pattern

# 1. Navigate to desired project
cd [project-name]/

# 2. Install dependencies
# Most Python projects here use uv:
uv sync
# Some projects use pip/requirements.txt:
# pip install -r requirements.txt

# 3. Configure environment
cp .env.example .env
# Edit .env with your API keys

# 4. Run the agent
# (varies by project - see individual READMEs)

Architecture Patterns

Model Context Protocol (MCP)

Multiple projects demonstrate different MCP implementation patterns:

• FastMCP ASGI: Direct FastAPI integration (mcp-financial, smithery-example)
• FastMCP Streamable HTTP: Modern bidirectional communication (fpl-deepagent)
• Bright Data MCP: External MCP server communication
• Notion MCP: Database and knowledge management integration

Agent Communication

• A2A Protocol: Remote agent-to-agent communication (agent2agent)
• State Management: Persistent conversation memory (bright-mcp-server-overview)

UI Integration Patterns

• React + ChatGPT: OpenAI Apps SDK integration (fpl-deepagent)
• Next.js + ChatKit: Agent Builder workflow integration (openai-chatkit-starter-app)
• Slack Bots: Event-driven chat interfaces (multiple projects)
• CLI Clients: Command-line agent interaction

Document Generation

• Claude Skills API: Automated document creation with streaming (claude-skills)
• Multi-Format Support: PowerPoint, Excel, Word, PDF generation
• Custom Skills: Uploadable skill packages for specialized tasks

Development & Testing

• MCP Playground: Development and testing environment (smithery-example)
• Multi-LLM Orchestration: Framework exploration (mastra-overview)
• Agent Builder: OpenAI workflow development (openai-chatkit-starter-app)

Integration Patterns

• Container Deployment: Docker and cloud-ready
• API Integration: RESTful agent endpoints
• Database Integration: Knowledge management systems
• Real-time Data: Live API integration with caching

Contributing

Each project welcomes contributions. Please:

1. Fork the repository
2. Create a feature branch
3. Follow the project's coding standards
4. Include tests where applicable
5. Submit a Pull Request

License

MIT License - see individual project LICENSE files for details.

Support & Resources

Documentation Links

• Model Context Protocol
• LangGraph Documentation
• OpenAI Agent SDK
• OpenAI Apps SDK
• OpenAI ChatKit
• OpenAI Agent Builder
• Claude Skills API
• Claude Programmatic Tool Calling
• Claude Tool Search
• Anthropic Blog - Advanced Tool Use
• Anthropic Blog - Code Execution
• Anthropic Console
• Google ADK
• FastMCP
• Mastra Framework
• Smithery
• Slack API

Platform-Specific Support

• Bright Data: brightdata.com/support
• Notion: developers.notion.com
• Fantasy Premier League: fpl.readthedocs.io
• Slack: api.slack.com/support

---

Built with ❤️ demonstrating the future of AI agent development