1st Place
dot-dash
We reimagined how you visualise and interact with your agents, by importing sessions from codex, and bringing these agents to life.
Core trick
Maps imported Codex session events to 3D desk or sofa states in real time through a relay server that streams updates to viewers.
Description
We reimagined how you visualise and interact with your agents, by importing sessions from codex, and bringing these agents to life. You can invite other collaborators to view your agents in their open office, view statistics such as context usage per agent, agent health, retries and errors. Click the first aid kit to run `codex doctor`. We also have a hammer, for when agents don't behave. This should fix most problems.
README excerpt
# Dots office A small 3D office for actual Codex sessions. Working agents sit at desks; completed or interrupted agents chill on sofas in the expanded lounge. New source work returns the same agent to its desk. Two clearly labelled animated showcase characters keep the room active by arriving, working, changing tasks, wandering, and leaving. They are separate from actual sessions and their counts. ## Run locally Node.js 22.13 or newer: ```sh npm install npm run build npm run dev:relay ``` In another terminal, run `npm run dev`, then open http://127.0.0.1:3000. The Vite frontend proxies the live relay to port 8787. The production worker serves both the built frontend and the same relay. `npm test` runs the source, room, validation, and behavior checks. ## Bring your agents and work together 1. Choose **Import agents**, enter your name, then **Create live room**. In an invitation, choose **Join with my Codex → Join this live room**. 2. Copy the live prompt into your own Codex. Choose an exact project and the actual session IDs you want to share. 3. Codex downloads the inspectable Node bridge, lists the selected project's sessions read-only, and starts the bridge after you choose the IDs. Keep that terminal open while working. 4. Use **Invite someone** to share the room link. Every visitor can join with their own selected sessions. Session IDs are scoped to each participant so contributors do not overwrite each other's agents. The bridge reads the local `state_5.sqlite`