FlyWhisperer

• GIF

  FlyWhisperer Demo

InspirationThe FlyWhisperer Elevator Pitch

try it online at https://flywhisperer.onrender.com/

The Problem "Imagine trying to navigate a maze of calendar widgets, ad pop-ups, and endless pricing filters using only a screen reader. For millions of visually impaired and elderly users, booking a simple flight online isn't just frustrating—it is often a barrier to independence."

The Solution "Enter FlyWhisperer. FlyWhisperer is a highly accessible, voice-first travel agent that completely bypasses visual UI clutter. Instead of clicking through screens, users simply press one large button and say: 'Find me the cheapest flight from Berlin to London next Friday.'"

The Tech Magic (WebMCP) "What makes FlyWhisperer unique is how it's built. Instead of a traditional standalone backend, it is built natively on the WebMCP standard and hosted via ChatGPT Sites. When the user speaks, the AI agent dynamically discovers our WebMCP tool, securely pulls real-time flight data, strips out all the marketing noise, and instantly reads back only the best option in clear, conversational audio—or displays it in ultra-high-contrast, WCAG AAA-compliant text."

The Impact "We aren't just building another travel app. We are showcasing how the WebMCP standard and AI can translate the web's most complex data into an inclusive, human-centric experience, giving visually impaired users their travel independence back."

What it does FlyWhisperer is an accessible, voice-first flight assistant designed for visually impaired and low-vision travelers. Instead of forcing users through cluttered booking forms and ads, it lets them simply speak or view ultra-high-contrast text to find and compare the cheapest flights effortlessly.

How we built it We built FlyWhisperer natively around the WebMCP standard and hosted it via ChatGPT Sites. The conversational agent leverages built-in Speech-to-Text and TTS capabilities to discover and invoke our custom WebMCP tool, which securely queries, parses, and sanitizes live flight data.

Challenges we ran into Handling large, complex flight payloads was our biggest hurdle. We had to strictly sanitize the data within our WebMCP worker to prevent prompt injection, preserve LLM token context, and translate technical airline jargon into natural, spoken English.

Accomplishments that we're proud of We successfully proved that WebMCP can transform an overwhelmingly visual, complex task into an intuitive, eyes-free experience that adheres strictly to WCAG AAA accessibility standards.

What we learned We gained deep insights into WebMCP tool discovery, edge tool calling, and how to design prompt boundaries specifically optimized for voice output and assistive technologies. videos are here: https://youtu.be/PmjaUKTGlQw https://youtu.be/FlIWOWxLpEA https://youtu.be/GZXBhJxg1j8

What's next for FlyWhisperer Next, we plan to support end-to-end booking execution directly through WebMCP tools, integrate real-time accessibility information for airports and airlines (e.g., wheelchair assistance, guide dog policies), and package it as an installable mobile PWA.

Built With

• aaa
• accessablity
• amadeus
• claude
• claudecode
• darkmodebydefault
• duffle
• inclusivedesign
• mcp
• node.js
• openai
• typescript
• voiceai
• wcag
• wcag-aaa
• webmcp
• webspeechapi