# hackshop (hackshop-mcp) > hackshop is an open-source MCP server and website that maps a project idea to hackable, repurposable, or protocol-native hardware — returning 3-5 candidates with brick-risk ratings, firmware links, live eBay searches, step-by-step how-tos, and AI-generated architecture diagrams. ## Product - [Home](https://www.hackshop.dev/): Live demo — describe a project idea and get hardware proposals with hack difficulty, brick risk, community size, used-market price ranges, and setup-time estimates. - [Templates](https://www.hackshop.dev/templates): Pre-filled project templates (old phone as a clock, e-paper calendar, Roomba as a robot, etc.). - [Inventory](https://www.hackshop.dev/inventory): Mark hardware you already own (stored in browser localStorage) so proposals can reuse it. - [Field guides](https://www.hackshop.dev/resources): Editorial guides on hacking and repurposing consumer hardware. ## MCP server hackshop is also an MCP (Model Context Protocol) server usable from Claude Desktop, Claude Code, Cursor, or any MCP-aware host: ```json { "mcpServers": { "hackshop": { "command": "npx", "args": ["-y", "hackshop-mcp"], "env": { "ANTHROPIC_API_KEY": "sk-ant-..." } } } } ``` - [Source on GitHub](https://github.com/msanchezgrice/hackshop-mcp): Open source, MIT-licensed. - [npm package](https://www.npmjs.com/package/hackshop-mcp): Published as `hackshop-mcp`. - [Suggest a device](https://github.com/msanchezgrice/hackshop-mcp/issues): Catalog contributions via GitHub issues. ## Agent-facing endpoints - [agents.md](https://www.hackshop.dev/agents.md): Onboarding and safe-action guide for browser agents. - [Agent card](https://www.hackshop.dev/.well-known/agent-card.json): A2A-style capability manifest. - [AI agent manifest](https://www.hackshop.dev/.well-known/ai-agent.json): Guardrails and permitted actions. ## API notes The live demo posts to `/api/propose` (rate limited to 5 requests/hour/IP; costs ≈ $0.002 per call). Other JSON endpoints: `/api/howto`, `/api/diagram`, `/api/assembly`, `/api/simulate`. All `/api/` routes are disallowed in robots.txt and intended for the site's own UI, not third-party crawling. ## Legal - [About](https://www.hackshop.dev/about) - [Contact](https://www.hackshop.dev/contact): msanchezgrice@gmail.com (include "Hackshop" in the subject line). - [Privacy](https://www.hackshop.dev/privacy) - [Terms](https://www.hackshop.dev/terms)