Claude Training Center
All programs

Building with the Claude API

Build a working application with the Claude API, from the first request to tool use, agents, evaluation and production concerns, including Arabic-language behavior.

What participants will be able to do

  • Call the Messages API, stream responses and get structured output
  • Give Claude tools, and connect systems through the Model Context Protocol
  • Build an agent with the Agent SDK, and decide when an agent is the wrong design
  • Write evaluations that catch regressions, including on Arabic text
  • Control cost and latency with model choice and prompt caching

What the program covers

  1. First calls

    Messages, system prompts, streaming and structured output.

  2. Tools and MCP

    Defining tools, handling results, and connecting to company systems.

  3. Agents

    The Agent SDK, agent loops, guardrails, and simpler designs that may serve better.

  4. Evaluation

    Building a test set from real inputs, in Arabic and English.

  5. Production

    Cost, latency, prompt caching, error handling and logging.

  6. Build lab

    Teams build a small feature against their own use case.

What you’ll have at the end

Each team leaves with a working prototype, an evaluation set, and a list of what must change before production.