Television gives an existing agent a place to put things that should outlive a chat response: status cards, documents, interactive views and web pages. Artifacts can be grouped into channels and changed later by asking the agent. It is a display and artifact workspace, not another inference provider or coding-agent orchestrator. The makers opened the repository to the public in early October; the npm package’s 1.4.23 release is dated October 2.

  • How it connects: A local server and tv CLI run on the same Linux or macOS machine as the agent. Installed skills teach the agent to create and update artifacts. A browser or the Apple Silicon Mac app is the viewer, and can run on a different machine.
  • Getting started: The README offers the Mac download and a Linux/browser setup guide. Its source-build path requires Node.js 24 and npm 11.5–11.x; the packaged CLI’s administration guide lists Node.js 22.12+ as its minimum. This is an extra service and skill bundle, not a feature already built into your agent.
  • Cost and access: Source is MIT-licensed; the repo and browser interface are available without an announced subscription. No claim here about future hosted pricing.
  • Caveats: The browser viewer is labeled experimental and cannot display external web pages inside an artifact like the Mac desktop app can. The privacy policy says the npm release sends content-free usage telemetry to US-hosted PostHog by default; DO_NOT_TRACK=1 before first use or tv telemetry disable opts out. Do not put the viewer on a public network without reviewing its token and network settings.

This is worth a look if your agent keeps producing throwaway local dashboards or useful output trapped in chat. It is based on the published repo and documentation, not a hands-on evaluation.