tropo
Documentation pages
On this page
  1. Steps
  2. Coming back to an agent later
  3. Common questions
  4. Related

Verified against v1.102.0 on 2026-09-30 by orpheus-o42

Create an agent

Ask Po for an agent that does one job. Three questions, one confirmation, about five minutes.

An agent in Tropo is a set of files in your Studio: who it is, what it is for, and what it has recorded. Because those are files, what it records in one session is there for the next, in whichever AI tool you open the Studio with. This page shows how to make one through Po, the concierge. You need a working Studio first; if you do not have one, start with the Quickstart.

If Po offered you Darin or Cal during setup and you said yes, you already have an agent. This page is for making your own.

Steps

  1. Open your Studio and tell Po. Open your AI tool in the Studio folder the way you did in the Quickstart. When Po greets you, say:

    I want to create my first agent.

  2. Answer three questions. Po asks them one at a time:

    • What kind of work should the agent handle?
    • What matters most in how that work gets done?
    • What is the first specific thing you would want it to do?

    Keep it to one job. If you ask for an agent that does everything, Po suggests starting with one thing; you can add more later.

  3. Confirm the proposal. Po shows you three lines and waits:

    • Name: a short name for the agent. Po suggests one that describes its job, such as researcher or drafter. Pick any name you like; your choice wins.
    • Role: one sentence in your own words.
    • First task: one concrete thing it will do in its first session.

    Say yes, or change anything. Nothing is written until you confirm.

  4. Po creates the agent. It takes a minute or two. Po tells you when it is done, starting with "Your agent is created and registered." She also tells you the agent now has a memory: your name, what you were working on and where you left off are saved in its files for the next session.

  5. Start the agent.

    1. Open a new session in your AI tool, in the same Studio folder.

    2. Attach the agent's activation file, agents/<name>/<name>-activation.md. For an agent named researcher that is agents/researcher/researcher-activation.md.

    3. Say, naming your agent:

      Activate Researcher using this file.

    4. When it has started, give it the first piece of work.

Coming back to an agent later

You do not go through Po again. In any future session, open the Studio folder, start a new chat, attach the agent's activation file, and send the same sentence:

Activate Researcher using this file.

Say it in words rather than relying on the attachment alone: not every AI tool treats an attached file as an instruction. Darin's and Cal's activation files also carry this sentence, with their own paths, under Come back tomorrow; an agent you create has no such section, so keep this page's sentence.

Common questions

  • Can I have more than one agent? Yes. Ask Po for another one the same way. If you want several agents that work together, tell Po you want to set up a team; she has a separate path for that.
  • Can I change what the agent does? Yes. Tell the agent itself. Asking for a whole new agent or a project sends you back to Po.
  • Where does the agent live? In the folder agents/<name>/ inside your Studio. Its memory is there too, not inside your AI tool, so if you open the Studio with a different tool, everything the agent saved comes with it. Conversation that was never saved to the Studio, and your AI tool's own settings, do not.