The latest articles on DEV Community by BeachFinder (@beachfinder). https://dev.to/beachfinder https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4039692%2F70a3734c-23fd-46af-a4bb-635ad55ce09d.png https://dev.to/beachfinder en BeachFinder Tue, 21 Jul 2026 09:12:36 +0000 https://dev.to/beachfinder/i-shipped-a-chatgpt-plugin-backed-by-a-remote-mcp-server-184900-beaches-live-data-one-backend-57eb https://dev.to/beachfinder/i-shipped-a-chatgpt-plugin-backed-by-a-remote-mcp-server-184900-beaches-live-data-one-backend-57eb <p>This week my ChatGPT plugin <strong>BeachFinder</strong> got approved and published. It lets ChatGPT answer questions like "warmest beach near Valencia right now?" or "compare three beaches near Miami for families" from live data — 184,900 real swim spots worldwide, with current sea temperature, waves, wind, UV and community reports.</p> <p>Here is the architecture decision that made it worth writing about: <strong>the plugin is just a thin skin over a remote MCP server.</strong></p> <h2> One backend, every assistant </h2> <p>The Model Context Protocol server lives at <code>getbeachfinder.com/mcp</code> (no auth, 12 read-only tools). The exact same endpoint serves:</p> <ul> <li>the ChatGPT plugin (<a href="https://chatgpt.com/plugins/plugin_asdk_app_6a55316d0c488191abdc21d086fdddb0" rel="noopener noreferrer">published this week</a>)</li> <li>Claude Desktop / Claude.ai connectors</li> <li>Cursor and any other MCP client</li> </ul> <p>So every new "AI app store" that appears is mostly a metadata submission, not new code. If you are building anything data-backed for AI assistants in 2026, I would start with the MCP server and treat each assistant surface as a distribution channel.</p> <h2> What I learned in review </h2> <ul> <li> <strong>Read-only tools sail through.</strong> Every tool is a search/read. No write, no auth, no user data. Review was smooth.</li> <li> <strong>Safety framing belongs in the tool descriptions.</strong> BeachFinder never declares a spot "safe to swim" — the tools say so explicitly, and defer to flags, lifeguards and local authorities. Reviewers notice.</li> <li> <strong>Example prompts are your listing's landing page.</strong> Ours map 1:1 to the strongest tools (conditions lookup, comparison, activity providers). Generic prompts waste the slot.</li> </ul> <h2> The data part </h2> <p>The hard work was never the plugin — it was the dataset: 184,900 deduplicated swim spots (OpenStreetMap + a long enrichment pipeline), live marine data via Open-Meteo, photos, community reports. The MCP layer is basically read-only RPC over that.</p> <p>Try it: search "BeachFinder" in ChatGPT plugins, or point any MCP client at <code>https://getbeachfinder.com/mcp</code> (docs at <a href="https://getbeachfinder.com/tools" rel="noopener noreferrer">getbeachfinder.com/tools</a>). Feedback very welcome — solo developer, I will take all of it.</p> architecture chatgpt mcp showdev