Build Screen Apps with
AI-Native from the Ground Up
The First Screen Platform Built for AI Agents
Most platforms bolt on AI for content generation. TelemetryOS enables AI agents to architect, implement, test, and deploy complete interactive applications — not just generate images.
- 18-Tool MCP Server
- A built-in Model Context Protocol server gives AI agents direct control over the development environment — store manipulation, screenshots, canvas control, log reading, and runtime management.
- 15 Domain-Specific Skills
- Structured skills encode deep platform expertise: requirements gathering, architecture patterns, UI design for signage and kiosks, data integration, testing, and debugging. Skills transform a general-purpose AI into a TelemetryOS specialist.
- Works with Any MCP Agent
- Claude Code, Cursor, Windsurf, or any MCP-compatible AI agent — zero configuration. Open a TelemetryOS project and the MCP connection, skills, and tool permissions are already set up. No vendor lock-in.
18 MCP Tools
The AI Agent Never Leaves the Editor
The MCP server gives the AI agent eyes (screenshots), hands (store and canvas control), and ears (log reading) — everything it needs to build autonomously.
- Store Tools — seed data, read state, debug. Six tools let the AI agent read, write, delete, and list keys across all four storage scopes. Seed test data, inspect application state, and reset to defaults — without leaving the code editor.
- Visual Capture — screenshots at every aspect ratio. The AI agent captures screenshots of the rendered application at any of 8 aspect ratio presets, from 5:1 ultra-wide to 1:5 portrait skyscraper. Automatic sweep mode captures all presets in one command for visual verification.
- Canvas Control — switch views, themes, backgrounds. Four tools control the development canvas: switch between render, settings, and web views, toggle light/dark mode, change background presets, and set aspect ratios. The agent sees exactly what the end user will see.
- Logs & Diagnostics — read errors, build output, platform logs. Three log tools surface console output, dev server messages, and platform operation logs with time and level filtering. The agent debugs issues by reading the same information a developer would see in DevTools.
Runtime & Configuration
- Runtime Control — reload and restart. Soft or hard reload the application, restart the development server. The agent can recover from errors and test fresh state without human intervention.
- Zero configuration required. Open a TelemetryOS project and everything is pre-configured — MCP connection, tool permissions, skill access, build commands. Zero permission prompts during normal AI-driven development.
15 Domain-Specific Skills
From General-Purpose AI to Screen App Specialist
Skills encode deep platform expertise into structured guidance that progresses logically — requirements feed architecture, architecture informs design, design guides implementation.
- Requirements Gathering
- A 6-phase interactive conversation (Vision, Render, Data, Settings, Multi-Mode, Summary) that produces a structured specification with store keys, data sources, and an implementation plan.
- Architecture & Design
- Mount point architecture, responsive scaling, signage display patterns, kiosk interaction patterns (touch, idle timeout, sessions), web view SPA routing, and settings component patterns.
- Data & Integration
- Store synchronization patterns between settings and render views, external API access via proxy fetch, weather API patterns, and media library access — the patterns that let applications connect to the systems you already run.
- Testing & Debugging
- Visual testing via MCP screenshot tools across all aspect ratios, store data testing via MCP store tools, and a common errors and fixes reference guide.
From Prompt to Production
The Session Ends with Screens Running Your App
The same workflow that built this menu board: a structured requirements conversation, skills-guided implementation, autonomous visual testing, and a publish step that puts the application on real screens.
- Requirements to Working App
- The agent scaffolds a project from a template, runs the structured requirements conversation, then implements store, settings, render, and web views with skills-guided code generation.
- Autonomous Visual Iteration
- Screenshot, read logs, adjust code, reload, screenshot again. The agent verifies its own work across every aspect ratio before a human ever reviews it.
- One-Click Publish
- The Developer App's publish action — version bump, cloud build, upload, and deploy to screens — is exposed through the MCP server, so the loop ends with content live in a real venue, not a local demo.
- Git-to-Screen CI/CD
- For ongoing development, connect GitHub and push a commit — TelemetryOS builds and deploys to every assigned screen automatically, with branch deployments, staged rollouts, and one-click rollback.
Complete AI Development Platform
Everything an AI Agent Needs
MCP tools, domain skills, visual testing, platform simulation, aspect ratio sweeps, and two paths to production — one-click publish and git-to-screen CI/CD
- MCP Server (18 Tools)
- Store, screenshot, canvas, log, runtime, and data tools — the AI agent never needs to leave the code editor.
- 15 AI Skills
- Requirements, architecture, design (signage, kiosk, web), data integration, testing, and debugging — structured platform expertise.
- Visual Feedback Loop
- AI takes screenshots, reads logs, adjusts code, and screenshots again — autonomous visual iteration without human intervention.
- Full Platform Simulation
- Storage, MQTT, weather, media, device capabilities — all simulated locally. Build and test without physical hardware.
- 8 Aspect Ratio Presets
- From 5:1 ultra-wide to 1:5 portrait skyscraper — AI tests every form factor automatically in one sweep.
- One-Click Publish
- Version bump, build, upload, and deploy to screens in one step — exposed to AI agents through the MCP server.
- Git-to-Screen CI/CD
- Push a commit and TelemetryOS builds and deploys to every assigned screen — branch deployments, staged rollouts, rollback.
- Agent-Agnostic
- Claude Code, Cursor, Windsurf, and any MCP-compatible agent work with zero configuration changes.
- Project Templates & CLAUDE.md
- Scaffold a working project in under a minute, with AI skills pre-loaded and a project configuration file that gives any agent session context and guardrails.
Go Deeper
Developer App →
Desktop IDE with live canvas preview, hot reload, and one-command deployment — home of the MCP server.
SDK →
React SDK with persistent storage, MQTT, hardware APIs, and 23 settings components.
Deployment Pipeline →
Push code, we build, host, and deploy to every screen — no infrastructure to manage.
Blog: Building Apps with AI Agents →
A walkthrough of the MCP workflow — from requirements conversation to deployed screen application.
Describe It. The AI Builds It.
From natural language requirements to a deployed screen application — with structured skills, visual testing, and one-command publishing.