AI EngineeringInterview preparation
Intermediate·Concept·2 min read

Microsoft Agent Framework

Practice this question

Beginner — What is MAF?

  • Microsoft Agent Framework provides agents and explicit workflows.
  • An agent handles model-driven interaction and tools; a workflow connects agents and functions through controlled execution paths.
  • It builds on ideas from AutoGen and Semantic Kernel.

Intermediate — When do you use an agent versus a workflow?

  • Use an agent for variable language interaction and bounded tool selection.
  • Use a workflow when order, dependencies, approvals, or state transitions must be explicit.
  • A business application often uses both.

Senior — MAF versus LangGraph?

  • I compare ecosystem/language fit, state contracts, review/resume behavior, persistence guarantees, observability, and hosting.
  • LangGraph starts from explicit state-graph orchestration; MAF provides agent and workflow abstractions in its ecosystem.
  • The validated operational contract matters more than branding.

Scenario — A client already uses Semantic Kernel. Do you rewrite?

  • I inventory maintained features, current support, and actual pain points.
  • I prototype MAF behind stable tool contracts and compare evaluation/recovery behavior before migrating.
  • A rename or new framework alone does not justify a risky rewrite.
Adapted from 13_AI.MD · “29.10 Microsoft Agent Framework”. Code examples are study fragments. Difficulty and core labels are editorial estimates; version-specific behavior should be checked against your project.