Claude & Anthropic Agentic Framework

Claude & Anthropic Agentic Framework

NEW

Deep dive into Claude's agentic capabilities — tool use, computer use, and the Agent SDK.

#technical AIClaudeAnthropic

Claude & Anthropic Agentic Framework

Concept & Motivation

Claude has become the reference model for agentic AI in enterprise settings. Its tool use capabilities, extended thinking, computer use, and the Agent SDK represent the most complete agentic framework available today. But the documentation is vast, the capabilities are evolving fast, and the gap between “works in a demo” and “works in production” is significant.

This course is a practitioner’s deep dive into building with Claude’s agentic stack — from Claude Code for rapid prototyping to the Agent SDK for production multi-agent systems.

What You’ll Learn

  • Claude’s agentic architecture — how tool use, extended thinking, and the instruction hierarchy work under the hood. The mental model that makes everything else click
  • Claude Code mastery — using Claude’s agentic coding tool for rapid prototyping, code generation, file manipulation, and system automation. CLAUDE.md configuration, hooks, and custom commands
  • Tool use design — designing tool schemas that Claude can reliably call. Input validation, error handling, and the difference between tools that work once and tools that work at scale
  • Extended thinking — when to enable thinking, how it affects output quality, and how to use thinking tokens for complex reasoning chains
  • The Agent SDK — building custom agent applications with Anthropic’s SDK. Agent loops, tool registration, state management, and multi-turn orchestration
  • Multi-agent patterns — supervisor-worker, pipeline, and specialized agent architectures using the SDK. When to use one capable agent vs. multiple specialized agents
  • Computer use — Claude’s ability to interact with GUIs. When it’s the right approach (legacy systems, no API) and when it’s the wrong one (anything with an API)
  • MCP (Model Context Protocol) — connecting Claude to external data sources and tools via the standardized protocol. Building and consuming MCP servers
  • Production deployment — rate limiting, error handling, retry strategies, prompt caching, and cost optimization. The engineering that turns a prototype into a reliable system

Who This Is For

  • Developers and engineers building agentic applications with Claude
  • Technical leads evaluating Claude for enterprise agent deployments
  • AI engineers comparing Claude’s agentic capabilities with alternatives
  • Product managers with technical background scoping Claude-powered features

Participants should be comfortable with APIs and basic programming. Python or TypeScript experience is helpful.

Format & Duration

2-day workshop (on-site). Day 1: Claude Code deep dive, tool use design, and extended thinking with guided exercises. Day 2: Agent SDK, multi-agent patterns, MCP integration, and a build session where participants create a production-ready agent for their use case.

What Makes This Course Different

This isn’t an API tutorial — the documentation already covers that. This course teaches the design decisions that the documentation doesn’t cover: when to use extended thinking vs. more tools, how to structure multi-agent systems that don’t spiral, how to write tool descriptions that work at the 99th percentile, and how to manage costs at scale.

The exercises use real business scenarios, not toy examples. And the patterns come from production deployments, including our own AgentForge platform which runs on Claude’s stack.

Q & A

Learn more about what we do

In production, model choice matters enormously. Claude's approach to tool use, instruction following, and safety is architecturally different from GPT or Gemini. Understanding one framework deeply gives you a reference point for evaluating all others. We chose Claude because its agentic capabilities are currently the most production-ready for enterprise use — but we compare with alternatives throughout.
Basic API familiarity helps. You should be comfortable reading JSON and understanding HTTP requests. The hands-on exercises use Claude Code (which requires no coding) and the Anthropic API (which requires basic scripting). We provide templates and pair technical with non-technical participants.
No. Claude Code is a standalone agentic coding tool that runs in your terminal and can read files, write code, and execute commands. The Anthropic API is the programmatic interface for building custom applications. This course covers both — Claude Code for rapid prototyping and the API + Agent SDK for production systems.
One module covers cost optimization: prompt caching, model selection (Haiku vs. Sonnet vs. Opus), token budgeting, and when to batch vs. stream. You'll build a cost model for your specific use case. Most participants are surprised that well-designed agent systems cost less than expected — it's the badly designed ones that burn budget.
Like

Get in touch
Published
On this page
Claude & Anthropic Agentic Framework Claude & Anthropic Agentic Framework Concept & Motivation What You’ll Learn Who This Is For Format & Duration What Makes This Course Different Q & A Learn more about what we do