Skybridge
A construction kit for the little interactive widgets you now see inside chat apps like Claude and ChatGPT — build one visual component once, and it works in every AI chatbot instead of you rebuilding it three times.
🔗 Visit Skybridge
Description
AI chat apps like Claude and ChatGPT increasingly let developers add small interactive components inside the conversation — think an embedded calendar picker or a live chart, not just text. Building one of those for each chat app separately is tedious, since each has its own quirks. Skybridge's whole job is to remove that duplication: write the interactive piece once, and it runs the same way across Claude, ChatGPT, VS Code, Cursor, and any other MCP-compatible client.
Skybridge is a free, open-source (MIT-licensed) full-stack TypeScript/React framework specifically for building MCP (Model Context Protocol) Apps — interactive widgets that render inside AI chatbots. It ships a local development emulator with hot-module reload, a built-in tunnel for testing without ngrok, type-safe end-to-end development in a tRPC-like style, and a compliance-audit tool for app-store-style submission requirements. It's built by Alpic, a Paris-based team that's an active contributor to the MCP specification and has been featured on OpenAI's developer blog; the GitHub repo (alpic-ai/skybridge) has around 1,945 stars, and the project claims 100K+ monthly npm downloads and roughly 10% market share of MCP apps running in Claude and ChatGPT — the market-share figure is self-reported and not independently verifiable.
💬 Our review
The short version: if you're building anything more interactive than plain text for Claude or ChatGPT's app ecosystems, Skybridge is a legitimate, free way to avoid writing separate implementations per platform — and being MIT-licensed with an active, credible team behind it (MCP spec contributors, OpenAI blog feature) makes it a safer open-source bet than a random weekend project.
There isn't a well-established direct competitor yet since MCP Apps as a category is brand new in 2026, which is both Skybridge's opportunity and its risk — being an early mover in a fast-moving spec means some churn is likely as the protocol itself evolves. Compared to hand-rolling your own MCP App integration per platform, Skybridge's write-once abstraction is clearly worth adopting unless you have a very platform-specific need; the main thing to watch is whether Alpic keeps pace as Anthropic, OpenAI, and others iterate on the underlying MCP Apps spec.
💰 Pricing
📊 Global score
🤖 AI-enriched data
Gratuit et open source (licence MIT) — pas de palier payant, monétisation possible côté hébergement/déploiement des apps construites avec
Pros
Gratuit, open source, licence MIT permissive
Écrire une fois, déployer sur Claude/ChatGPT/VS Code/Cursor
Émulateur local avec hot-reload + tunnel intégré pour tester sans ngrok
Équipe crédible : contributeurs actifs de la spec MCP, cités par le blog développeurs d'OpenAI
Cons
Chiffre de '~10% de part de marché des MCP Apps' auto-déclaré, non vérifié
Catégorie MCP Apps très jeune — risque de changements de spec impactant le framework
Pas encore de concurrent direct établi pour comparer objectivement