Termi Protocol
Watch your AI coding agents build, live in 3D
294 followers
Watch your AI coding agents build, live in 3D
294 followers
The Termi Protocol is a 3D simulation of AI agent workflows. Give your coding agents a face, a desk and a living room. Watch them read, write and run commands live in 3D, like a game. You run the agents; we visualize the process.
Katalyst
Will watching them not burn more tokens? and be shit more expensive?
Termi Protocol
@divyansh_lohia
No, quite the opposite; our goal beyond just gamification is to ensure that every time you assign a new agent, the model added to the project consumes fewer tokens. We achieve this through a history and graph system, where the graph operates based on "hive mind" dynamics.
That's interesting. Does it work with any agent that exposes terminal output, or only specific ones?
Termi Protocol
@dhiraj_patel5 Currently, there is support for seven or eight CLI agents—including Claude, Codex, X.AI, GitHub Copilot, Amazon Q, Aider, and OpenCode allowing you to connect all your agents and seamlessly create and manage workflows.
Congrats on the launch! Any plans to visualize non coding agents too? Such as research or marketing agents?
Termi Protocol
@betsy_seus
Thank you so much for the support!
For software and coding agents, our visualization support is very strong, and this is currently our main focus. I’d say we support that side at 100%.
For non coding agents such as research or marketing agents, we currently have only a very minimal level of support. They can create basic tasks and simple notifications, but I’d say this is still around 10% of where we want it to be.
Would you personally like to see stronger visualization support for research, marketing, or other non coding agents? If there is enough demand, I’d be happy to include this more seriously in our development roadmap.
Thanks again for the thoughtful question.
Congrats with a launch!
So we're not all confused here, it's just a fun project, you are not suggesting it as a serious workflow, right?
Termi Protocol
@valzubkov
Thank you, I really appreciate it.
This project is our main ship. Termi is planned as a 9 stage roadmap, and we want to improve it step by step instead of trying to force everything into the first version.
The current stage is what we call the Milano Protocol. Its focus is turning the agent workflow into a 3D and gamified experience. The next stage will be the Istanbul Protocol, where we will focus more on identifying missing parts, improving workflow control, and making agent management easier in real use.
So yes, the playful side is intentional, but this is also a long-term product for making coding agent work more understandable and manageable.
Everyone is asking how the room shows agent state - the state I care most about is mine. Running several agents, the biggest silent time sink is agents sitting blocked on my approval while I look somewhere else. The room shows their work; the killer view would be my inbox: one glance at everything currently waiting on me, oldest first. Is there anything like that today, or is approval surfaced per agent only? Congrats on the launch!
the visualization solves the "is my agent working or hallucinating" problem instantly. i've been staring at scrolling logs for weeks and this is basically the diff.
real q: when claude and codex share a repo, is there a way for the human (or a third agent) to sign off on which commits actually shipped good work vs which shipped garbage? that trust signal would compound over time and route future tasks to whichever agent is best at what.
The state I can never eyeball fast in scrolling logs is idle-waiting-on-approval versus stuck-in-a-retry-loop. They throw off similar output volume but mean opposite things, and I usually notice the difference three minutes too late. If the room made those two visually distinct at a glance, an agent standing still at the desk versus pacing and redoing, that's the thing that would earn the screen space over an 8 hour session for me. Does the body reflect that an agent is actually blocked, or just that it's between commands?