# Termi Protocol shows coding agents working in 3D and hands off stalled tasks

published: 2026-10-23T16:05:59.452Z
topic: infrastructure
source: https://www.producthunt.com/products/termi-protocol
url: https://aiagentnews.io/stories/termi-protocol-shows-coding-agents-working-in-3d-and-hands-off-stalled-tasks

Termi Protocol is a desktop app that shows AI coding agents such as Claude Code, Codex, Gemini and OpenCode working at desks in a 3D room, with each file read, command and edit displayed as it happens. Its latest update lets a session stopped by a usage, rate or context limit pass to an agent on another model, and it now runs on macOS, Windows and Linux.

Termi, made by Eric Omer Ercan, displays real CLI coding agents as figures working at their own desks in a 3D room. Users watch each file read, command and edit as it happens, and risky actions wait for their approval. The maker says Termi is not a chat app or an IDE and works with the agents users already run.

The headline change is limit handoff. When an agent hits a usage, rate or context limit, Termi saves the goal, changed files, recent commands, errors, open todos and the conversation. The user picks an agent on any model and chooses how much context to send, and nothing is sent until they confirm. Claude Code reports the stop through its hooks, while other CLIs are read from the terminal.

Other additions include a boss agent that splits one goal across idle agents, a trash that restores deleted agents with their memory and tasks, and a Files and Source Control sidebar for committing inside the app. A browser demo is available without installing anything.

## Why it matters if you build agents

If you run several coding agents, a stalled task can move to another model without being explained again from scratch. Seeing edits and commands as they happen gives you a point to catch problems before risky actions run. Termi sits on top of CLI agents you already use, so check which of your agents it can read before relying on handoff.