AI NewsDev toolsAnnouncement

Telepath released Television, an MIT-licensed app that gives a coding agent a canvas for documents, charts and live web pages next to the chat

Telepath released Television, an MIT-licensed sidecar app that gives a chat agent a visual canvas for artifacts. The agent writes documents, data views, visualisations and live web pages into named channels on the TV, where they stay in view across the whole session. It works with Hermes, OpenClaw, Pi, Claude Code and Codex, and the GitHub repository reached 118 stars in three days.

AI News

Editorial3 min read

LinkedInX
GitHub social card for the telepath-computer/television repository

Image: GitHub

Why it mattersAn agent that drops its output on a canvas the user can see all at once is a different reading surface from a scrolling chat, and makes a long session easier to review without losing earlier results.

A chat with a coding agent scrolls one answer at a time, so an answer from an hour ago is already off the screen when the next one lands. Telepath's Television, released on 2026-10-01, gives the agent a second screen for the parts of the answer that should stay in view.

Television is an MIT-licensed sidecar app for a chat agent, published at telepath-computer/television on GitHub. The repository reached 118 stars in three days. Telepath, the publisher, describes itself as a software company working on interfaces for personal computers. The licence credits Unternet PBC.

What it does

Telepath says Television "gives you a visual space for creating and working with artifacts in collaboration with your agent." Instead of returning its answer inside the chat window, the agent can write an artifact onto the TV and modify it later. The project lists documents, data, visualisations, live web pages and small apps as the kinds of artifacts the agent can place there. Artifacts are persistent, so a document the agent drafted this morning is still on the TV when the user comes back to it in the afternoon, and the agent can edit it in place.

The user keeps using the same chat interface they already use. The phrase Telepath suggests is "put that on my TV", which tells the agent to render its output on the TV. The documentation says the agent will often decide on its own when a response belongs on the TV.

How the three parts fit together

Television has three components. A server runs on the same machine as the agent and is the piece the agent talks to. A skills bundle is installed during setup; Telepath says it "teaches your agent how TV works, how to create artifacts on your TV, and how to modify the artifacts and the TV environment." A client app is the TV itself: an installable Electron app on macOS, or a browser-based interface on any other platform. Telepath says the Electron build has extra abilities, including loading external web pages inside an artifact.

Television works with any agent harness that supports skills and file access, and Telepath specifically names Hermes, OpenClaw, Pi, Claude Code and Codex. The agent itself must run on Linux or macOS; agents running on Windows are not supported. Building Television from source needs Node.js 24 and a specific range of npm.

Why it matters

A reader looking at a long chat with a coding agent has already lost the first half of the session to scrolling. The information that would be easiest to re-read, like a list of files the agent has edited, a diagram of the data it pulled, or the UI it is working on, has to be re-generated or re-opened each time it is asked for. A canvas the user can glance at while reading the chat keeps those pieces alive for the length of the session, which is what Telepath is calling an interface for personal computing with an agent inside it.

The spec-driven development process is only open to Telepath for now, so the project will take feedback and bug reports on GitHub but not outside pull requests yet. The Discord link in the README is the route for anyone who wants to help.

Source

Primary source: telepath-computer/television on GitHub. Publisher: Telepath at telepath.computer.

SourceTelepath

This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.

Share
LinkedInX