Welcome to drio

drio is a visual builder that lets anyone create and deploy MCP apps to ChatGPT, Claude, and Cursor — without writing server code. You design tool UIs, wire up API connections, and ship from a single canvas. What previously required a specialized engineering team and weeks of work now takes an afternoon.

Something big is happening with how people use software. AI assistants are becoming the primary way millions of people interact with digital services. Over 1.2 billion people use AI clients monthly — ChatGPT, Claude, Cursor, Copilot, and dozens more. And they are not just chatting. They are searching for products, querying databases, managing projects, and running business workflows — all inside the conversation.

The protocol making this possible is MCP (Model Context Protocol), an open standard created by Anthropic and now supported by OpenAI, Google, Microsoft, JetBrains, and 70+ AI clients. MCP lets AI assistants connect to external tools and data — and render interactive UIs directly in the chat. If you want to understand the protocol in depth, we wrote a What Is MCP? explainer.

But building MCP apps? That has been painful. Until now.

What you can build

With drio, the entire process becomes visual:

• Tool builder — Define your tool's inputs, logic, and outputs on a flow-based canvas. Each tool is a discrete capability the AI can invoke — search products, check order status, generate reports.

• Widget designer — Pick from 21 built-in widget primitives and customize them visually. Tables, charts, forms, maps, product cards, carousels, stat dashboards — all render inline in the chat. Check out our guide on Building MCP Tools with Rich UIs for the full breakdown.

• API connector — Connect to any REST API with our visual request builder. URL templates with variable interpolation, custom headers, query params, auth handling — all configured visually. No HTTP client code to write.

• One-click deploy — Ship your MCP server to the cloud and connect it to any AI assistant instantly. Your tool gets a dedicated endpoint at /{your-app}/mcp that works across all supported clients.

A few things people are building

The range of what is possible is wider than you might expect:

• E-commerce assistants — Product search with card carousels, cart management, and order tracking, all inside ChatGPT
• Internal dashboards — Sales teams querying CRM data, marketing teams viewing campaign analytics, all without leaving Claude
• Customer support tools — Interactive troubleshooting flows with form widgets for data collection and stat cards for account status
• Data explorers — SQL query tools that render results as interactive charts and tables inside any AI client

We ship 9 production-ready templates covering these verticals and more — e-commerce, healthcare, legal, productivity, analytics, and food & beverage. Each template includes pre-configured API connections, widget assignments, and branding presets.

How it works

The workflow is straightforward. Four steps, and you are live.

1. Design your tool on the drio canvas — drag widgets, define input parameters, map data fields
2. Connect to external APIs using the visual request builder with built-in auth handling
3. Preview your tool in the built-in chat playground before going live — you see exactly what users will see
4. Deploy with one click — drio compiles your design into an optimized, spec-compliant MCP server

Something worth noting: drio is not generating code with AI. It is a compiler architecture. Your visual configuration gets compiled into a typesafe, deterministic MCP application. No inference, no hallucinated code, no "AI magic." The output is always correct because it is compiled, not generated.

Updates are instant too. Save in the builder, and changes are live on the MCP endpoint within seconds. No redeploy, no build step.

Why this matters now

We are in the early days of a platform shift. Websites gave way to mobile. Mobile is now giving way to AI clients as the dominant interface between people and software.

The Model Context Protocol is the standard that makes this shift practical. Anthropic open-sourced it in November 2024, and adoption has been staggering — the reference server repository has over 80,000 GitHub stars, SDKs exist for 10 programming languages, and every major AI company has committed to support it.

But the tooling for actually building MCP apps has lagged far behind. Most teams still need to write TypeScript, understand the protocol specification, manage server infrastructure, handle OAuth, build widget UIs in React, and ensure compatibility across dozens of AI clients. That is a $200K+/year engineering investment. We cover this in more detail in Why We Built drio.

drio eliminates that barrier. If you can use a visual builder, you can build an MCP app.

Get started

drio is currently in early access. Join the waitlist to be among the first to build MCP apps visually. We are rolling out invites every week.

If you are already in the program, start with the quickstart guide — you can have a tool live in ChatGPT in under 10 minutes. Or check out Connecting Your First API to learn how the API connector works.

Our goal is simple: anyone with a great idea for an AI tool should be able to build and ship it in an afternoon.