Choosing a Framework

• Key points are in the visual diagram above.

The 2026 landscape is noisy. Interviews and production reviews both ask why this runtime, not a list of logos you installed.

You do not pick a database by GitHub stars. You pick Postgres vs a graph DB by the queries. Same for agents.

Scenario

Ticket router with human approval → LangGraph. Five specialists writing a report → CrewAI. Internal Python API with strict JSON → PydanticAI. GPT-only prototype this week → OpenAI Agents SDK.

Solution

In Agent Framework Landscape, apply Choosing a Framework to this scenario: Ticket router with human approval → LangGraph. Identify the inputs, run the technique, validate the output, and note one thing you would monitor in production.

Do this before moving to the next module — reading alone is not enough.

Open the Code Walkthrough below and run it locally. Change one parameter related to Choosing a Framework (e.g. model, temperature, top_k, or tool name), observe the difference in output, and write 2–3 sentences explaining what changed.

Highlighted lines show where Choosing a Framework happens in the code.

1def pick_framework(needs):  # define a reusable function2    if needs["hitl"] or needs["checkpoints"] or needs["branching"]:3        return "langgraph"  # return the result4    if needs["roles"] and needs["team"]:5        return "crewai"  # return the result6    if needs["typed_python"] and needs["structured_out"]:7        return "pydantic-ai"  # return the result8    if needs["openai_only"] and needs["handoffs"]:9        return "openai-agents"  # return the result10    if needs["coding_agent"] or needs["claude_code"]:11        return "claude-agent-sdk"  # return the result12    if needs["microsoft_stack"]:13        return "microsoft-agent-framework"  # return the result14    if needs["gemini_native"]:15        return "google-adk"  # return the result16    return "raw loop is still allowed"  # return the result

• Running two orchestrators in one service
• Picking a framework before writing the state machine on paper

Quick recap — the most important points from this module.