← Back to pathways

From Prompts to Agents

Build AI systems that use tools, search for information, and complete multi-step tasks autonomously.

Advanced4 stages10 recipes17-25 hours

This advanced pathway covers the frontier of practical AI: giving language models the ability to use tools and work autonomously on complex tasks.

We start with tool use - teaching AI to call functions, query databases, and interact with external systems. Then we move to agents - AI systems that can plan, execute, and iterate on multi-step workflows without human intervention.

These techniques are powerful but require careful implementation. We focus on practical patterns that work reliably in production, not experimental approaches that fail unpredictably.

This pathway assumes you are comfortable with Python, APIs, and have completed the "API Loops" pathway or have equivalent experience.

Before you start

  • Completed "API Loops and Programmatic AI" pathway or equivalent experience
  • Comfortable with Python including async/await patterns
  • Understanding of REST APIs and JSON
  • Experience with at least one LLM API (OpenAI, Anthropic)
  • Willingness to debug and iterate on complex systems

What you will achieve

  • Build AI systems that use external tools and data sources
  • Create autonomous workflows that complete multi-step tasks
  • Deploy user-facing AI applications
  • Understand the current frontier and limitations of AI agents

Pathway stages

1

Tool Use Foundations

4-6 hours

Give AI the ability to call functions, query data, and interact with external systems.

Recipes in this stage

What you will be able to do

  • Implement function calling with Claude or GPT
  • Build AI assistants that query live data
  • Combine multiple tools in a single conversation
2

Autonomous Workflows

4-6 hours

Chain AI capabilities together into workflows that run without human intervention.

Recipes in this stage

What you will be able to do

  • Build end-to-end automated workflows
  • Chain different AI models for different tasks
  • Handle errors and edge cases in autonomous systems
3

User-Facing Applications

5-7 hours

Build AI-powered applications that your team or service users can interact with.

Recipes in this stage

What you will be able to do

  • Deploy AI interfaces for non-technical users
  • Ground AI responses in your actual content
  • Build reliable production systems
4

Advanced Patterns

4-6 hours

Sophisticated techniques for complex real-world problems.

Recipes in this stage

What you will be able to do

  • Handle complex multi-document analysis
  • Use AI for strategic and analytical thinking
  • Build predictive systems that inform decisions

Created by AI Recipes Team

Last updated: 2024-12-23