Overview

This repository contains workshop materials from Anthropic's "Code with Claude" sessions. It provides hands-on guidance and practical examples for leveraging Claude's capabilities, particularly within the realm of managed agents and advanced AI workflows. While not actively maintained or accepting contributions, these resources offer valuable insights into building sophisticated AI applications.

Key Features

• Managed Agent Architectures: Explore comprehensive examples of designing, composing, and shipping Claude Managed Agents for complex tasks.
• Agent Decomposition and Orchestration: Learn to break down intricate problems into smaller, manageable skills and orchestrate them using Claude's MCP framework.
• Eval-Driven Development: Discover methodologies for iterative agent development where evaluations and metrics guide prompt engineering and configuration.
• Memory and Persistence: Implement strategies for endowing agents with memory capabilities, enabling them to retain context across sessions and consolidate information.
• AI-Assisted Workflows: Walk through end-to-end AI-assisted product development and research processes, from interview to deployment.
• Model and Parameter Optimization: Gain practical experience in selecting optimal Claude models and inference parameters for cost-effectiveness and performance.

Typical Use Cases

• Building Advanced Conversational AI: Develop sophisticated agents capable of complex reasoning, multi-step task execution, and maintaining context over extended interactions.
• Automating Research and Analysis: Create agents that can ingest and analyze large datasets, research SEC filings, or perform due diligence for M&A activities.
• Streamlining Development Workflows: Utilize AI to assist in code generation, design exploration, and runtime verification of application components.
• Developing Intelligent Tools and Dashboards: Construct interactive applications, such as incident dashboards, powered by AI agents that can interact with local tools and logs.
• Experimenting with Agent Competition: Engage in scenarios where agents are configured to compete or collaborate in simulated environments, optimizing for specific objectives.
• Educational Resource for Claude Developers: Ideal for developers looking to deepen their understanding of Claude's managed agent capabilities and best practices.