If you’ve been following AI development closely, you’ve started seeing MCP mentioned everywhere \u2014 in Claude’s settings, in developer blogs, in job descriptions. Most explanations are either too abstract or too deep in the weeds.<\/p>\n
This is the explanation I wish I’d had when I first encountered it.<\/p>\n
Imagine you’re a developer who wants Claude to help you manage your GitHub issues, update your Jira tickets, query your database, and send Slack messages \u2014 all from a single conversation.<\/p>\n
Before MCP, every one of these integrations was a custom engineering project. You’d need to:<\/p>\n
It was like the early days of computing, when every printer needed its own custom driver for every operating system. Chaos.<\/p>\n
MCP is the USB standard for AI integrations. Build it once, works with any MCP-compatible model.<\/p>\n
Model Context Protocol is an open standard \u2014 published by Anthropic in November 2024, now adopted by OpenAI, Google, and most major AI tooling companies \u2014 that defines a universal way for AI models to connect to external tools and data sources.<\/p>\n
At its core, MCP has three pieces:<\/p>\n
MCP Servers<\/strong> \u2014 programs that expose tools and data to AI models. A GitHub MCP server exposes tools like “list my repositories”, “read a file”, “create a pull request”. A database MCP server exposes “run a query”, “show table schema”. You can build an MCP server for anything.<\/p>\n MCP Clients<\/strong> \u2014 AI applications that connect to MCP servers. Claude Desktop is an MCP client. Any application built on Claude’s API can be an MCP client. The client asks the server “what tools do you have?” and uses them during conversations.<\/p>\n MCP Protocol<\/strong> \u2014 the communication standard between clients and servers. JSON-RPC based, runs over stdio or HTTP with SSE. Simple enough that building a basic MCP server takes about an hour.<\/p>\n Here’s what happens when you ask Claude “what GitHub issues are assigned to me?” with a GitHub MCP server connected:<\/p>\n From your perspective: you asked a question and got an answer. Under the hood: an entire API integration happened invisibly.<\/p>\n That’s a real, working MCP server. Claude can now call “get_project_status” during any conversation where this server is connected.<\/p>\n You don’t have to build everything from scratch. The MCP ecosystem has grown rapidly:<\/p>\n Restart Claude Desktop and you have GitHub, PostgreSQL, and file system access in every conversation.<\/p>\n Here’s a practical example \u2014 an MCP server that lets Claude query and manage your Laravel application:<\/p>\n Add this to your Claude Desktop config and you can now say:<\/p>\n And Claude does it \u2014 directly, no copy-pasting, no tab switching.<\/p>\n If you’ve used the OpenAI API, you’ve used function calling. MCP and function calling solve similar problems \u2014 letting AI use tools \u2014 but at different levels:<\/p>\n Think of function calling as a walkie-talkie \u2014 works great for one conversation. MCP is a phone system \u2014 infrastructure that handles all conversations.<\/p>\n MCP’s real significance isn’t developer productivity \u2014 it’s what it enables for AI agents. An AI agent with MCP access to your entire business toolstack can:<\/p>\n Without a human in the loop. Without custom code connecting each step. That’s not a future possibility \u2014 it’s happening with MCP today.<\/p>\nHow It Works in Practice<\/h2>\n
\n
The MCP Server Architecture<\/h2>\n
\/\/ A minimal MCP server in TypeScript\r\nimport { Server } from \"@modelcontextprotocol\/sdk\/server\/index.js\";\r\nimport { StdioServerTransport } from \"@modelcontextprotocol\/sdk\/server\/stdio.js\";\r\n\r\nconst server = new Server(\r\n { name: \"softcrony-tools\", version: \"1.0.0\" },\r\n { capabilities: { tools: {} } }\r\n);\r\n\r\n\/\/ Define available tools\r\nserver.setRequestHandler(\"tools\/list\", async () => ({\r\n tools: [\r\n {\r\n name: \"get_project_status\",\r\n description: \"Get the current status of a client project\",\r\n inputSchema: {\r\n type: \"object\",\r\n properties: {\r\n project_id: {\r\n type: \"number\",\r\n description: \"The project ID\"\r\n }\r\n },\r\n required: [\"project_id\"]\r\n }\r\n }\r\n ]\r\n}));\r\n\r\n\/\/ Handle tool calls\r\nserver.setRequestHandler(\"tools\/call\", async (request) => {\r\n if (request.params.name === \"get_project_status\") {\r\n const projectId = request.params.arguments.project_id;\r\n\r\n \/\/ Call your actual API\r\n const response = await fetch(`https:\/\/api.softcrony.com\/projects\/${projectId}`);\r\n const project = await response.json();\r\n\r\n return {\r\n content: [\r\n {\r\n type: \"text\",\r\n text: JSON.stringify(project, null, 2)\r\n }\r\n ]\r\n };\r\n }\r\n});\r\n\r\n\/\/ Start server\r\nconst transport = new StdioServerTransport();\r\nawait server.connect(transport);<\/code><\/pre>\nMCP Servers That Already Exist<\/h2>\n
\n\n
\n \nCategory<\/th>\n Available MCP Servers<\/th>\n<\/tr>\n<\/thead>\n \n Code & Version Control<\/td>\n GitHub, GitLab, Bitbucket<\/td>\n<\/tr>\n \n Project Management<\/td>\n Jira, Linear, Asana, Trello, Notion<\/td>\n<\/tr>\n \n Communication<\/td>\n Slack, Discord, Microsoft Teams<\/td>\n<\/tr>\n \n Databases<\/td>\n PostgreSQL, MySQL, SQLite, MongoDB, Redis<\/td>\n<\/tr>\n \n Cloud<\/td>\n AWS, Google Cloud, Cloudflare<\/td>\n<\/tr>\n \n Observability<\/td>\n Datadog, Grafana, Sentry<\/td>\n<\/tr>\n \n Browser<\/td>\n Puppeteer, Playwright, Browserbase<\/td>\n<\/tr>\n \n File System<\/td>\n Local files, Google Drive, Dropbox<\/td>\n<\/tr>\n \n APIs<\/td>\n Stripe, Twilio, SendGrid, HubSpot<\/td>\n<\/tr>\n \n Search<\/td>\n Brave Search, Tavily, Exa<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n Setting Up MCP in Claude Desktop<\/h2>\n
\/\/ ~\/.claude\/claude_desktop_config.json\r\n{\r\n \"mcpServers\": {\r\n \"github\": {\r\n \"command\": \"npx\",\r\n \"args\": [\"-y\", \"@modelcontextprotocol\/server-github\"],\r\n \"env\": {\r\n \"GITHUB_PERSONAL_ACCESS_TOKEN\": \"your-token-here\"\r\n }\r\n },\r\n \"postgres\": {\r\n \"command\": \"npx\",\r\n \"args\": [\"-y\", \"@modelcontextprotocol\/server-postgres\"],\r\n \"env\": {\r\n \"POSTGRES_CONNECTION_STRING\": \"postgresql:\/\/user:pass@localhost\/db\"\r\n }\r\n },\r\n \"filesystem\": {\r\n \"command\": \"npx\",\r\n \"args\": [\r\n \"-y\",\r\n \"@modelcontextprotocol\/server-filesystem\",\r\n \"\/Users\/yourname\/projects\"\r\n ]\r\n }\r\n }\r\n}<\/code><\/pre>\nBuilding a Custom MCP Server for Your Laravel App<\/h2>\n
\/\/ mcp-server\/index.ts\r\nimport { Server } from \"@modelcontextprotocol\/sdk\/server\/index.js\";\r\nimport { StdioServerTransport } from \"@modelcontextprotocol\/sdk\/server\/stdio.js\";\r\n\r\nconst API_BASE = process.env.LARAVEL_API_URL;\r\nconst API_TOKEN = process.env.LARAVEL_API_TOKEN;\r\n\r\nconst api = async (endpoint: string, method = \"GET\", body?: object) => {\r\n const response = await fetch(`${API_BASE}${endpoint}`, {\r\n method,\r\n headers: {\r\n \"Authorization\": `Bearer ${API_TOKEN}`,\r\n \"Content-Type\": \"application\/json\",\r\n },\r\n body: body ? JSON.stringify(body) : undefined,\r\n });\r\n return response.json();\r\n};\r\n\r\nconst server = new Server(\r\n { name: \"laravel-app\", version: \"1.0.0\" },\r\n { capabilities: { tools: {} } }\r\n);\r\n\r\nserver.setRequestHandler(\"tools\/list\", async () => ({\r\n tools: [\r\n {\r\n name: \"list_users\",\r\n description: \"List all users in the application\",\r\n inputSchema: {\r\n type: \"object\",\r\n properties: {\r\n page: { type: \"number\" },\r\n search: { type: \"string\" }\r\n }\r\n }\r\n },\r\n {\r\n name: \"get_order_stats\",\r\n description: \"Get order statistics for a date range\",\r\n inputSchema: {\r\n type: \"object\",\r\n properties: {\r\n from: { type: \"string\", description: \"Start date YYYY-MM-DD\" },\r\n to: { type: \"string\", description: \"End date YYYY-MM-DD\" }\r\n },\r\n required: [\"from\", \"to\"]\r\n }\r\n },\r\n {\r\n name: \"run_artisan\",\r\n description: \"Run a safe Laravel artisan command\",\r\n inputSchema: {\r\n type: \"object\",\r\n properties: {\r\n command: {\r\n type: \"string\",\r\n enum: [\"cache:clear\", \"queue:restart\", \"optimize\", \"migrate:status\"]\r\n }\r\n },\r\n required: [\"command\"]\r\n }\r\n }\r\n ]\r\n}));\r\n\r\nserver.setRequestHandler(\"tools\/call\", async (request) => {\r\n const { name, arguments: args } = request.params;\r\n\r\n switch (name) {\r\n case \"list_users\":\r\n const users = await api(`\/api\/users?page=${args.page ?? 1}&search=${args.search ?? \"\"}`);\r\n return { content: [{ type: \"text\", text: JSON.stringify(users, null, 2) }] };\r\n\r\n case \"get_order_stats\":\r\n const stats = await api(`\/api\/stats\/orders?from=${args.from}&to=${args.to}`);\r\n return { content: [{ type: \"text\", text: JSON.stringify(stats, null, 2) }] };\r\n\r\n case \"run_artisan\":\r\n const result = await api(\"\/api\/artisan\", \"POST\", { command: args.command });\r\n return { content: [{ type: \"text\", text: result.output }] };\r\n\r\n default:\r\n throw new Error(`Unknown tool: ${name}`);\r\n }\r\n});\r\n\r\nconst transport = new StdioServerTransport();\r\nawait server.connect(transport);<\/code><\/pre>\n\n
MCP vs Function Calling \u2014 What’s the Difference?<\/h2>\n
\n\n
\n \n<\/th>\n Function Calling<\/th>\n MCP<\/th>\n<\/tr>\n<\/thead>\n \n Scope<\/td>\n Per-API-call<\/td>\n Persistent server<\/td>\n<\/tr>\n \n Setup<\/td>\n Define in each API request<\/td>\n Configure once, use everywhere<\/td>\n<\/tr>\n \n Portability<\/td>\n Provider-specific<\/td>\n Works with any MCP-compatible model<\/td>\n<\/tr>\n \n State<\/td>\n Stateless<\/td>\n Can maintain state across calls<\/td>\n<\/tr>\n \n Authentication<\/td>\n Manual per call<\/td>\n Handled by server<\/td>\n<\/tr>\n \n Best for<\/td>\n Single API integrations<\/td>\n Full tool ecosystem<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n Why MCP Matters Beyond Developer Tools<\/h2>\n
\n