

Establish a problem-first, evidence-based method for defining framework requirements, classifying the changing landscape, and deciding whether an agent framework is justified at all.
Module objectives
- Convert a realistic use case into measurable framework requirements and explicit disqualifiers.
- Classify a current offering by category, ownership boundary, runtime, and deployment model using official evidence.
- Construct and justify a weighted selection scorecard without turning it into a popularity ranking.
- Identify use cases in which a direct model SDK, deterministic workflow, or custom application loop is simpler and more maintainable.
Detailed topic content is generated and researched separately so each topic maintains its own source trail.
This module in 6 slides
Click through for a quick overview, or use your arrow keys.
Put this module to work
Suggested time: 3 to 4 hours
Real-world scenario
A startup picks a popular agent framework because it trended online. Three months later, they need durable workflows that pause for approvals and resume days later, and their framework is not built for that. Picking a framework is an architecture decision. This module helps you match requirements to frameworks before you write a line of code.
Step-by-step walkthrough
- Write your requirements: execution style, state and persistence, human-in-the-loop, tool integrations, observability, deployment target, language and team skills.
- Mark each requirement as must-have, nice-to-have or not needed.
- Shortlist three frameworks (for example LangGraph, Google ADK, PydanticAI, CrewAI or a plain SDK) and check each against the must-haves using current documentation.
- Note maturity signals: release cadence, breaking changes, documentation quality and community.
- Write a one-page fit assessment with your provisional choice and open questions.
Common mistakes and how to fix them
I have built with many tools over 25 years, and the most expensive mistakes were always picking a platform before writing down requirements. Frameworks are no different. Spend an afternoon on the requirements list and you can save months.
Try this
Ask yourself: if this framework disappeared next year, how hard would it be to move? Write down the parts of your design that would need rewriting.
Before you move on
Your checklist is saved in this browser.
Agent opportunity scorer
Rate your idea on five factors and get a recommendation: agent, plain workflow, or simplify first. Your work is saved in this browser.
Autonomy envelope builder
List the actions your agent might take and sort each into Do alone, Ask first or Never. Your work is saved in this browser.