--- title: AI Travel Concierge - Multi-Agent MCP System emoji: โœˆ๏ธ colorFrom: indigo colorTo: purple sdk: gradio sdk_version: "5.9.1" app_file: app.py pinned: true license: mit tags: - mcp-in-action-track-consumer - mcp - agents - travel - ai-assistant - multi-agent - gradio-6 - flux - modal - image-generation short_description: "8 AI Agents plan your perfect trip with real booking links" --- # โœˆ๏ธ AI Travel Concierge - Multi-Agent MCP System > **๐Ÿ† MCP Hackathon Winter 2025 Submission** > **Track 2: MCP in Action - Consumer Category**
![AI Travel Concierge Banner](https://img.shields.io/badge/๐Ÿค–_8_AI_Agents-Working_Together-blue?style=for-the-badge) ![MCP Powered](https://img.shields.io/badge/MCP-Powered-green?style=for-the-badge) ![Gradio 6](https://img.shields.io/badge/Gradio-6.0-orange?style=for-the-badge) **Your Personal AI Travel Planning Team** ๐ŸŒ [โ–ถ๏ธ Watch Demo Video](#demo-video) | [๐Ÿš€ Try Live Demo](https://huggingface.co/spaces/MCP-1st-Birthday/AI-Travel-Concierge)
--- ## ๐ŸŽฏ What is AI Travel Concierge? AI Travel Concierge is a **complete multi-agent AI system** that plans your entire trip using **8 specialized MCP servers** working together. Unlike simple chatbots, this system uses **autonomous AI agents** that: - ๐Ÿง  **Reason & Plan** - Understands your travel preferences and creates personalized itineraries - ๐Ÿ”ง **Use Real Tools** - Each agent has specialized tools via MCP (Model Context Protocol) - ๐Ÿ”— **Generate Real Booking Links** - Every recommendation includes working links to Google Flights, Booking.com, Viator, and more - ๐ŸŽจ **Create Visual Content** - Generates custom AI travel posters using [FLUX.2-dev](https://huggingface.co/black-forest-labs/FLUX.2-dev) on Modal --- ## โœจ Key Features ### ๐Ÿค– 8 Specialized AI Agents | Agent | MCP Server | Capabilities | |-------|------------|--------------| | ๐ŸŒค๏ธ **Weather Agent** | `weather_server.py` | Real-time forecasts, best time to visit, packing suggestions | | โœˆ๏ธ **Flights Agent** | `flights_server.py` | Flight search with Google Flights & Skyscanner links | | ๐Ÿจ **Hotels Agent** | `hotels_server.py` | Luxury hotel recommendations with Booking.com links | | ๐ŸŽฏ **Activities Agent** | `activities_server.py` | Tours, attractions with Viator & GetYourGuide links | | ๐Ÿฝ๏ธ **Dining Agent** | `dining_server.py` | Restaurant recommendations with OpenTable links | | ๐Ÿš— **Transport Agent** | `transport_server.py` | Airport transfers, car rentals with real provider links | | ๐Ÿงญ **Recommendations Agent** | `recommendations_server.py` | Smart destination finder for undecided travelers | | ๐ŸŽจ **Poster Agent** | `poster_server.py` | AI-generated travel posters via Modal [FLUX.2-dev](https://huggingface.co/black-forest-labs/FLUX.2-dev) | ### ๐ŸŽฏ Smart Destination Finder **Don't know where to go?** Our intelligent recommendation system: - Analyzes your interests (beach, adventure, culture, food, etc.) - Considers your budget and travel dates - Shows current deals and seasonal discounts - Scores destinations to find your perfect match ### ๐Ÿ’Ž Premium Glass UI - Beautiful dark mode glass-morphism design - Mobile-responsive layout - Real-time agent status indicators - One-click quick destination buttons ### ๐Ÿ”— Real Booking Links Every recommendation includes **working booking links**: - โœˆ๏ธ Google Flights & Skyscanner for flights - ๐Ÿจ Booking.com & Hotels.com for accommodations - ๐ŸŽซ Viator & GetYourGuide for activities - ๐Ÿฝ๏ธ TripAdvisor & OpenTable for restaurants - ๐Ÿš— Blacklane & Welcome Pickups for transfers --- ## ๐Ÿ› ๏ธ Technical Architecture ``` โ”Œโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ” โ”‚ GRADIO 6 FRONTEND โ”‚ โ”‚ (Premium Glass UI + Chatbot) โ”‚ โ””โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”ฌโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”˜ โ”‚ โ–ผ โ”Œโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ” โ”‚ AI ORCHESTRATOR (Qwen 235B) โ”‚ โ”‚ Autonomous Planning, Reasoning, Execution โ”‚ โ””โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”ฌโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”˜ โ”‚ MCP Protocol โ”Œโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”ผโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ” โ–ผ โ–ผ โ–ผ โ”Œโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ” โ”Œโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ” โ”Œโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ” โ”‚ Weather โ”‚ โ”‚ Flights โ”‚ โ”‚ Hotels โ”‚ โ”‚ MCP โ”‚ โ”‚ MCP โ”‚ โ”‚ MCP โ”‚ โ””โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”˜ โ””โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”˜ โ””โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”˜ โ”‚ โ”‚ โ”‚ โ–ผ โ–ผ โ–ผ โ”Œโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ” โ”Œโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ” โ”Œโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ” โ”‚Activity โ”‚ โ”‚ Dining โ”‚ โ”‚Transportโ”‚ โ”‚ MCP โ”‚ โ”‚ MCP โ”‚ โ”‚ MCP โ”‚ โ””โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”˜ โ””โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”˜ โ””โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”˜ โ”‚ โ”‚ โ–ผ โ–ผ โ”Œโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ” โ”Œโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ” โ”‚ Recom- โ”‚ โ”‚ Poster โ”‚ โ”‚mendationโ”‚ โ”‚ MCP โ”‚โ”€โ”€โ–บ Modal FLUX.2-dev โ””โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”˜ โ””โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”˜ ``` ### Tech Stack - **Frontend**: Gradio 6 with custom CSS (glass-morphism) - **AI Model**: Qwen/Qwen3-235B-A22B-Instruct-2507 (Nebius) - **MCP**: Model Context Protocol for tool integration - **Image Gen**: Modal [FLUX.2-dev](https://huggingface.co/black-forest-labs/FLUX.2-dev) MCP for travel posters - **Protocol**: Server-Sent Events (SSE) for real-time streaming --- ## ๐Ÿš€ Quick Start ### Try it Online ๐Ÿ‘‰ [Launch AI Travel Concierge on Hugging Face Spaces](https://huggingface.co/spaces/ABDALLALSWAITI/AI-Travel-Concierge) ### Run Locally ```bash # Clone the repository git clone https://huggingface.co/spaces/ABDALLALSWAITI/AI-Travel-Concierge # Install dependencies pip install -r requirements.txt # Set your API key export NEBIUS_API_KEY="your-api-key" # Run the app python app.py ``` --- ## ๐Ÿ“น Demo Video > **[๐Ÿ“บ Watch Full Demo Video (YouTube Link)](#)** > > *Coming soon - 3-5 minute walkthrough showing:* > - Natural language trip planning > - All 8 agents in action > - Real booking link generation > - AI poster creation > - Smart destination finder --- ## ๐ŸŽฎ How to Use ### Option 1: I Know Where I Want to Go Simply type your trip request: > *"Plan a romantic trip from New York to Paris for 2 people, December 15-22, moderate budget, interested in culture and food"* The AI will automatically: 1. ๐ŸŒค๏ธ Check weather and best time to visit 2. โœˆ๏ธ Find flight options with booking links 3. ๐Ÿจ Recommend hotels with prices and links 4. ๐ŸŽฏ Suggest activities and tours 5. ๐Ÿฝ๏ธ Find the best restaurants 6. ๐Ÿš— Arrange airport transfers 7. ๐ŸŽจ Generate a custom travel poster! ### Option 2: Help Me Decide Click **"๐Ÿค” Find My Perfect Destination"** and tell us: - Your interests (beach, adventure, culture, etc.) - Your budget range - When you want to travel We'll match you with the **perfect destinations** and show current deals! --- ## ๐Ÿ“ Project Structure ``` ai-travel-concierge/ โ”œโ”€โ”€ app.py # Main Gradio application โ”œโ”€โ”€ weather_server.py # Weather MCP server โ”œโ”€โ”€ flights_server.py # Flights MCP server โ”œโ”€โ”€ hotels_server.py # Hotels MCP server โ”œโ”€โ”€ activities_server.py # Activities MCP server โ”œโ”€โ”€ dining_server.py # Dining MCP server โ”œโ”€โ”€ transport_server.py # Transport MCP server โ”œโ”€โ”€ recommendations_server.py # Smart recommendations MCP server โ”œโ”€โ”€ poster_server.py # AI poster generation MCP server โ”œโ”€โ”€ requirements.txt # Python dependencies โ””โ”€โ”€ README.md # This file ``` --- ## ๐ŸŒŸ What Makes This Special? | Feature | Traditional Chatbots | AI Travel Concierge | |---------|---------------------|---------------------| | Architecture | Single LLM | 8 Specialized MCP Agents | | Tool Usage | Basic API calls | Full MCP Protocol | | Booking Links | Generic or none | Real, working links | | Personalization | Limited | Smart matching algorithm | | Visual Output | Text only | AI-generated posters | | Agent Autonomy | None | Full planning & reasoning | --- ## ๐Ÿ‘จโ€๐Ÿ’ป Team **Solo Developer**: [@ABDALLALSWAITI](https://huggingface.co/ABDALLALSWAITI) --- ## ๐Ÿ“ฑ Social Media > **[๐Ÿฆ View on X/Twitter](#)** - *Link to social post* > > **[๐Ÿ’ผ View on LinkedIn](#)** - *Link to social post* --- ## ๐Ÿ™ Acknowledgments - **Anthropic** - For creating MCP and co-hosting the hackathon - **Gradio Team** - For Gradio 6 and organizing this amazing event - **Nebius** - For providing Qwen API credits - **Modal** - For Flux image generation MCP --- ## ๐ŸŽจ The Flux Journey: From Training to Deployment ### Custom LoRA Training for This Hackathon For this hackathon, I trained a custom **LoRA model specifically for travel posters** using [FLUX.1-Krea-dev](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev): ๐Ÿ”— **[Trip Poster LoRA Model on CivitAI](https://civitai.green/models/2146545/trip-poster)** This LoRA was trained to generate professional travel agency style posters with: - Bold destination typography - Polaroid-style photo collages - Travel-themed visual elements - Professional marketing aesthetics ### Pivoting to FLUX.2-dev Initially, I planned to use [FLUX.1-Krea-dev](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev) with my custom LoRA. However, when **[FLUX.2-dev](https://huggingface.co/black-forest-labs/FLUX.2-dev)** was released with significant improvements, I decided to pivot and spent **2 days** deploying it on Modal instead. ### Deep Study: FLUX.2-dev Prompt Engineering for Tourism Posters I conducted extensive research on FLUX.2-dev's prompt techniques to craft the perfect prompts for professional tourism posters. Key findings: - **Structured prompt format** - FLUX.2-dev responds best to detailed, well-organized prompts - **Style keywords** - Terms like "travel agency poster", "wanderlust aesthetic", "professional marketing" trigger the right visual style - **Typography handling** - Specific instructions for bold destination text placement - **Composition cues** - Polaroid collages, landmark silhouettes, gradient backgrounds - **Color psychology** - Warm tones for beach destinations, cool tones for winter escapes This prompt engineering research ensures the Poster Agent generates stunning, professional-quality travel posters every time. I documented the entire deployment process: ๐Ÿ“ **[LinkedIn Post: Deploying Flux MCP on Modal](https://www.linkedin.com/posts/abdallah-issac_mcp-ai-machinelearning-activity-7400147561761697792-TFkL?utm_source=share&utm_medium=member_desktop&rcm=ACoAABflfdMBdk1lkzfz3zMDwvFhp3Iiz_I4vAw)** ๐Ÿ“– **[Complete Deployment Guide (GitHub Gist)](https://gist.github.com/al-swaiti/8ea1d7df6ab728d111406b70e26150be)** The guide covers: - Setting up Modal for GPU inference - Deploying [FLUX.2-dev](https://huggingface.co/black-forest-labs/FLUX.2-dev) with optimal settings - Creating an MCP server interface - Integrating with Gradio applications - Performance optimization tips This enables the **Poster Agent** to generate stunning AI travel posters in ~30 seconds using H100 GPUs on Modal with the latest FLUX.2-dev model. --- ## ๐Ÿ“„ License MIT License - Feel free to use and modify! ---
**Built with โค๏ธ for the MCP Hackathon Winter 2025** [![Hugging Face](https://img.shields.io/badge/๐Ÿค—_Hugging_Face-Space-yellow?style=for-the-badge)](https://huggingface.co/spaces/ABDALLALSWAITI/AI-Travel-Concierge)