A Dispatch, start to finish
You describe an outcome. The Project Manager turns it into a Dispatch, proposes the Worker for the station, and sets it moving. The Developer writes the code and opens a pull request on the repository you authorised. Quality reads the work against the brief: it passes the Dispatch to merge, or returns it with notes and the Developer goes again. You watch all of it on the Floor, in real time, and you decide what merges.
The work is real
This is the part most tools skip. The Developer is an AI coding agent that commits to your repository, against your branch protections, and opens a pull request you review and merge like any other. There is no transcript to copy by hand and no code to paste. What lands in your repo is the product.
Where the Workers run
Each Worker runs in its own sandbox, with access only to the Git remote you authorise. It clones the repository, works on a branch, runs the suite, and opens the pull request. Nothing touches your machine: no Docker to run, no daemon to install, no server for you to keep alive. The software automation is hosted, so you open a browser and the office is already lit.
What you control
You are the CEO. You choose the work and the expected output, set hard spending caps, and decide what merges. Engage the Workers you need; dismiss the ones you do not, at any hour, with no notice and no severance. The Ledger records every Dispatch, line by line.
How this is different
Most AI agent products are management layers for engineers: you bring your own agents, run your own server, write the wiring, then watch a dashboard. The Floor is the opposite. The Workers come included, the office is hosted, and the proof is a real pull request, not a row in a table. If you have looked at agent frameworks or self-hosted task boards and wanted the company rather than the toolkit, this is that.