aws_mcp_hackathon

# Prosthetic Design  [](https://devpost.com/software/t-e5rljs) 🏆 **Winner of the Vizcom: Visual Intelligence Award** at the [MCP and A2A Hackathon - AWS Edition](https://lu.ma/vibecode?tk=V8VB1G) Personalized prosthetics, powered by style. Design a prosthetic leg that feels like **you**. --- ## Inspiration Our inspiration came from a close friend who needed a prosthetic leg. He wanted something that made him feel happy and reflected his personality, but he struggled to find a company that offered full customization. That experience made us realize how limited the options are for personalized prosthetics, and we wanted to build something to change that. ## What it does The **Prosthetic Design** web application lets users select stylistic preferences—​color schemes, patterns, textures, or themes—​through a simple dropdown interface. These preferences are used to generate a one-of-a-kind prosthetic leg design, which is then forwarded directly to a prosthetics manufacturer for production. ## How it works 1. **Vizcom automation** – Vizcom's design AI has no public API, so we built an `Apify` **Actor** (see `my-actor/src/main.py`) that drives Vizcom through a headless **Playwright** browser session. The Actor uploads a source image *or* a text-prompt, waits for Vizcom to generate a 3-D model, and automatically exports a `GLB`/`STL` file. 2. **MCP orchestration** – Mastra **MCP** coordinates multiple agentic components, invoking the Apify Actor and post-processing the result. 3. **Arcade.dev e-mail automation** – When a design is ready, an automated e-mail with the 3-D model attached is sent to the selected prosthetics manufacturer. ``` ┌─────────────┐ 1 ┌──────────────┐ 2 ┌───────────────┐ 3 ┌────────────────┐ │ Frontend │──────────▶│ Apify Actor │──────────▶│ MCP Flow │──────────▶│ Arcade.dev │ │ (Next.js) │ prompt │ (Playwright)│ .glb/.stl │ (TypeScript) │ email │ Manufacturer │ └─────────────┘ └──────────────┘ └───────────────┘ └────────────────┘ ``` ## System architecture  ### Vizcom automation demo  *The Apify Actor (agent) controls Vizcom headlessly via Playwright, performing all clicks, uploads, and exports without human intervention.* ## Tech stack - **Python** + **Playwright** + **Apify SDK** for browser automation - **Mastra MCP** for orchestration / agent framework - **Arcade.dev** for transactional e-mail - **Node.js / TypeScript / Next.js** for the web UI - **Vizcom** generative 3-D design engine ## Repository structure ``` . ├── README.md ← you are here ├── my-actor/ ← Apify Actor (Python + Playwright) │ ├── src/main.py ← main automation script │ ├── requirements.txt ← Python dependencies │ └── ... ├── frontend/ ← (optional) Next.js web app — not included in this repo └── ... ``` ## Getting started (local) ### Prerequisites - Python ≥ 3.10 - Node.js ≥ 18 (if you plan to run the frontend) - [Apify CLI](https://docs.apify.com/cli) `npm i -g apify` - Playwright browsers: `python -m playwright install firefox` ### 1 · Clone & install ```bash git clone https://github.com/your-org/prosthetic-design.git cd prosthetic-design/my-actor pip install -r requirements.txt ``` ### 2 · Configure secrets Create a `.env` file (or export env-vars) with your Vizcom credentials: ```dotenv VISCOM_USER=you@example.com VISCOM_PASSWORD=super-secret ``` ### 3 · Run the Actor ```bash # from the my-actor directory apify run --debug --headed # headed mode helps when debugging the browser ``` The Actor will upload `leg.png` by default (or download an image when `--imageUrl` is supplied), generate a 3-D model, and export `exported.glb` to the Apify key-value store. ### Optional CLI flags ```bash --imageUrl <url> # remote image to use instead of leg.png --prompt "neon pink cyberpunk pattern" # combine prompt + image workflow --headed # run browser in non-headless mode --debug # verbose logging ``` ## Deploy to Apify Cloud ```bash # authenticate once apify login # paste your API token # deploy actor from my-actor directory apify push ``` Then trigger the Actor either via the Apify UI or a simple HTTPS request: ```bash curl -X POST \ -d '{"imageUrl":"https://example.com/leg.jpg"}' \ "https://api.apify.com/v2/acts/<username>~prosthetic-design/run-sync?token=<APIFY_TOKEN>" ``` ## Challenges - **No Vizcom API** – we reverse-engineered the UI and built resilient Playwright selectors with retry logic. - **Headless browser quirks** – WebGL and dynamic React components required multiple workarounds. ## Accomplishments - Fully automated Vizcom generation flow with zero manual clicks - Seamless e-mail hand-off to manufacturers - End-to-end personalization from user style input to production-ready 3-D model ## What we learned - Deep dive into Apify's Python SDK and storage primitives - Coordinating multiple agentic frameworks (Apify, MCP) in a single product ## Sample generated models  _{Four color-way examples} ## Roadmap - Expand beyond legs: arms, feet, and limb covers - ✅ Add live 3-D preview in the web UI (three.js) - Integrate pricing & order-tracking APIs from manufacturers ## License Distributed under the MIT License. See [`LICENSE`](LICENSE) for more information.