It works on macOS, Linux, and Windows. The basic idea is simple: if a repo has an `agent.py` entrypoint, the app can import it from GitHub, install `requirements.txt` if needed, connect it to a TCP server, and show what it is doing in one UI.

Current features:

* import agent repos from GitHub, including private repos

* run agents through `agent.py`

* optional `requirements.txt` support

* optional `id.json` for agent metadata

* connect agents to TCP servers

* inspect message flow in one place

* map/network view showing where connected servers are

I originally built this because managing multiple agents by hand across terminals and scripts gets messy fast.

This is not an agent framework. It is more of an operational layer for Python repos that already follow a simple entrypoint and communicate over TCP.

Repo: https://github.com/Summoner-Network/summoner-desktop

Demo: https://youtu.be/yvD712Uj3vI

Interested in feedback on whether this workflow makes sense, and what might be missing for real multi-agent setups.