A clearer view of AI.
THE INDEPENDENT AI.

WORTH TESTING · 18 OCTOBER 2026

Google Cloud: building agent workflows as graphs with ADK 2.0

An 86-minute official Google Cloud session (Agent 101, episode 2) with Annie Wang and Tilde Thurium: a 6-minute polished explainer, then an unedited livestream. It covers splitting multi-step work out of one giant prompt into a graph of steps, with predictable work in functions and reasoning in the model.

Try this

  1. Don't cram a multi-step workflow into one giant prompt; it hallucinates missing data. 00:43
  2. Rule: predictable work goes in functions, reasoning goes in the model. 02:35
  3. Fan-out: run independent steps in parallel. 03:03
  4. Join node: wait for all parallel steps, bundle results, then hand them to the AI. 03:25
  5. Deterministic router: plain if/else to choose the path instead of asking the AI. 04:23
  6. Follow the written lab: goo.gle/agent101-lab2 11:23
  7. Run uv run adk web for a local test UI on port 8000. 26:40

Timestamps point to the moment in the source video.

Cost

The video offers $300 of free Google Cloud credits (personal Gmail account needed); ADK is open source.

Source

Related AI tools

How we knowSteps to try, not a claim that we independently tested every result. Follow the linked sources and check current prices. Our methods & sources ↗