TELEVISION Goes Open Source: The Model-Agnostic Agent Canvas and Personal Computer Interface
The TELEVISION team open-sourced their agent-native desktop interface, offering an open alternative to proprietary canvases like ChatGPT Space that connects directly to local models, Anthropic Claude, and OpenAI APIs.
The development team behind TELEVISION released the platform's complete codebase to the open-source community. Billed as the next iteration of the personal computer interface, TELEVISION replaces static chat lists with a persistent, spatial workspace where agents and users co-create documents, codebases, and interactive simulations.
The project emerges as an open-source answer to proprietary desktop environments such as OpenAI's ChatGPT Space, giving users full ownership over their workspace layouts, local documents, and model routing.
Architectural Contrast: Proprietary Chat vs. TELEVISION Spatial Computing
Traditional conversational interfaces lock user data inside linear scrolling timelines. When an agent creates five files, users must hunt through previous turns to recover code snippets.
TELEVISION shifts the interaction model to a persistent spatial canvas:
| Capability | ChatGPT Space (Closed SaaS) | TELEVISION (Open Source) |
|---|---|---|
| Model Compatibility | OpenAI models only | Model-agnostic (Claude, Gemini, DeepSeek, Local Ollama) |
| Source Availability | Proprietary closed-source | 100% Open source on GitHub |
| Data Locality | Cloud-hosted on OpenAI servers | Stored on local disk in JSON and Markdown formats |
| Extensibility | Closed platform extensions | Native Model Context Protocol (MCP) and custom React tiles |
| Desktop Performance | Web wrapper (Chromium) | Lightweight Tauri + Rust backend (sub-90MB memory) |
| Offline Capability | Non-functional without internet | Full offline functionality with local weights |
The Anatomy of an Agent Tile
Inside TELEVISION, content exists as modular tiles arranged across an infinite two-dimensional grid. Agents and human operators interact with three primary tile primitives:
┌────────────────────────────────────────────────────────┐
│ TELEVISION Spatial Canvas │
├────────────────────┬──────────────────┬────────────────┤
│ [Code Tile] │ [Web Preview] │ [Agent Log] │
│ src/server.ts │ localhost:3000 │ Tool Calls: │
│ • Syntax highlight│ • Hot reloading │ 1. read_file │
│ • AST diffing │ • DOM inspection│ 2. run_tests │
├────────────────────┴──────────────────┴────────────────┤
│ [Prompt Dock & Dispatcher] │
└────────────────────────────────────────────────────────┘
- Document & Code Tiles: Reactive editors powered by Monaco and CodeMirror. When an agent suggests refactors, diffs render inline side-by-side with acceptance toggles.
- Runtime Preview Tiles: Sandboxed iframes capable of mounting local Vite dev servers, Next.js instances, or raw SVG charts directly inside the canvas.
- Execution & Trace Tiles: Real-time terminal streams displaying agent reasoning traces, command line outputs, and file system mutations.
Quickstart: Cloning and Launching TELEVISION
TELEVISION ships as a cross-platform desktop binary for macOS, Linux, and Windows. Developers can build from source using Bun and Rust:
# 1. Clone the repository
git clone https://github.com/television-app/television.git
cd television
# 2. Install dependencies and compile the Rust core
bun install
bun run tauri dev
Once compiled, users can open the Settings pane and register their API credentials or configure an unauthenticated local endpoint pointing to an Ollama server running at http://localhost:11434.
Integrating Custom Models and MCP Tools
TELEVISION uses standard JSON configuration files to define model providers and MCP connectors:
{
"activeWorkspace": "project-nebula",
"providers": [
{
"id": "anthropic",
"model": "claude-sonnet-5.5",
"apiKeyEnv": "ANTHROPIC_API_KEY"
},
{
"id": "ollama-local",
"model": "deepseek-coder-v2:16b",
"endpoint": "http://localhost:11434/v1"
}
],
"mcpServers": {
"git": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-git"]
},
"filesystem": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-filesystem", "/Users/dev/projects"]
}
}
}
Why the Community is Moving Toward Agent-Native Desktops
The release of TELEVISION reflects a broader transition in developer workflows. The separation between text editors, terminal emulators, and chat interfaces creates unnecessary cognitive drag.
By unifying code editing, visual rendering, and autonomous agent loops into a single spatial canvas that runs offline and respects user data sovereignty, TELEVISION establishes an open foundation for modern personal computing.