The Project Agents Starter Kit

What this solves

Structure for more than one agent at once.

This kit assumes you already have files that persist and a way to keep them current. What is left is structure: one agent per project instead of one agent for everything, a coordinator that routes between them, model routing so cost matches the work, review that runs only where being wrong is expensive, and a board that shows you all of it at once.

It needs a platform you can configure: something that supports more than one running context, a way to pass instructions between them, and a place to persist records outside of any single conversation. Start with The Files-First Starter Kit and The Context Capture Starter Kit first if you have not built those yet.

The five pieces

Five pieces, built after the files already work.

The basic workflow

Build it in this order.

1. Give each project its own agent. Its own memory, task records, and states, so work never bleeds between projects.
2. Add a coordinator. Routes a request to the right project and never does the work itself.
3. Route by model. Match model cost and capability to the task's category and complexity.
4. Scale review. Run an independent pass only where being wrong is expensive: code and anything going outside.
5. Build the board. One view across every project, with whatever is blocked on you pulled to the top.

What you end up with

One system, even with several agents running.

• Each project runs from its own context, undiluted by the others
• A single place to ask what is in motion, across every project
• Cost that matches the task, not a flat rate for everything
• Review that catches real mistakes without slowing down routine work