Building with Google's Agent Development Kit (ADK)

The agent loop you keep rewriting is already built, and ADK is Google's version of it. Building with Google's Agent Development Kit (ADK) takes the lid off that loop and builds production agents on top of it.

Across eleven parts it covers the loop itself, why tool descriptions make or break an agent, multi-agent teams, session state and memory, streaming so your user does not think it crashed, keeping sensitive instructions away from the model, growing an agent's powers in the order you will want them, long-running execution, stateful deployment, and catching silent regressions.

Start at Part 1, or jump to any part below.

The parts

  1. The Agent Loop You Keep Rewriting Is Already Built
  2. The Magic in Your ADK Agent Is a Loop. Here Is the Loop.
  3. Your ADK Agent Is Only as Good as Its Worst Tool Description
  4. Stop Cramming One Agent With Ten Jobs. Build a Team.
  5. Your ADK Agent Forgot the User's Name. Here Is the Prefix That Would Have Saved It.
  6. Your ADK Agent Goes Silent for Forty Seconds. Your User Thinks It Crashed.
  7. The One Instruction You Never Want Reaching the Model, and the Six Lines That Stop It
  8. Three Ways to Give Your ADK Agent New Powers, in the Order You Will Want Them
  9. Your Forty-Minute Agent Died at Minute Thirty-Nine. One Config Flag Would Have Saved the Other Thirty-Eight.
  10. You Cannot Ship a Stateful Agent Like a Stateless Lambda
  11. Your Agent Worked in the Demo and Silently Regressed Three Commits Later