Agent patterns

Agent pattern decks

Short visual guides to designing AI agents people can actually trust. Each one takes a single problem, a handover, a plan preview, an agent register, and works it through in ten or so slides you can read in a couple of minutes.

They draw on the agent patterns library, a set of forty-one interaction patterns for agent products. Free to download, nothing to sign up for.

  1. Eight ways agent products lose trust

    Failure patterns that turn up across shipped products and the 2026 research, and the one change that does most: make the agent hand back receipts.

    PDF11 slides

    Download
  2. The six stages of agent handover

    Brief, plan, run, check, verify, learn. What each stage is for, and which two are the ones people notice are missing.

    PDF12 slides

    Download
  3. Five minutes before the agent starts

    What to agree before an agent acts: the plan preview in plain words, and the four screens that sit on top of it.

    PDF10 slides

    Download
  4. Reading a crowd of agents

    What a board of parallel runs has to show, and why the one row waiting on a person belongs at the top of it.

    PDF10 slides

    Download
  5. Your newest teammate has a spend cap

    An agent register: the owner, who may task it, what it is allowed to write, and the cap. The line everything else hangs off.

    PDF10 slides

    Download
  6. Can an agent use your product?

    What an agent actually reads on a checkout page, and why publishing an llms.txt is the cheapest way to find out who is already arriving that way.

    PDF10 slides

    Download