Anthropic Courses: Working with Claude — From API to Tools

Topics covered

• •Get an API key and make your first request to Claude through the Python SDK.
• •Master the messages format and the `user`/`assistant` roles, including the prefill technique.
• •Navigate the Claude model family and choose a model deliberately.

Topics covered

• •Master the basic structure of a correct prompt and the system prompt.
• •Learn to formulate clear and direct instructions.
• •Apply role prompting to change the tone and quality of responses.

Topics covered

• •Apply step-by-step thinking (chain-of-thought / "precognition") for complex tasks.
• •Use few-shot examples to set format and behavior.
• •Reduce hallucinations: the right to say "I don't know", evidence quotes, temperature 0.

Topics covered

• •Systematize the techniques covered and understand when to apply each one.
• •Break down the "anatomy" of large production prompts using real cases.
• •Master the prompt engineering lifecycle: from draft to iterative improvement.

Topics covered

• •Understand why prompt evaluations matter and what an eval consists of (input, reference, grader).
• •Distinguish grading types: human, code-based, and model-graded.
• •Write code-based and classification evals in plain Python.

Topics covered

• •Understand the four-step tool use cycle and the role of client code in it.
• •Define tools: `name`, `description`, `input_schema` (JSON Schema).
• •Get structured JSON from the model via a "forced" tool call.