{"id":"what-is-an-mcp-server","slug":"what-is-an-mcp-server","title":"What Is an MCP Server? The Plain-English Definition","excerpt":"An MCP server exposes your tools and data to AI models like Claude, ChatGPT, and Cursor through a standardized protocol. Here's exactly what one is, how it works, and why every modern SaaS ships one.","content":"<p><strong>Direct answer:</strong> An <strong>MCP server</strong> is a small program that exposes a set of tools, resources, and prompts to AI models through the Model Context Protocol. It's the bridge between an AI client (Claude, ChatGPT, Cursor) and a real-world system (your CRM, your database, your file system). If MCP is the standard, the MCP server is the <em>plug</em>.</p>\n\n<h2>Definition</h2>\n<blockquote>\n<p><strong>MCP Server:</strong> A program that implements the Model Context Protocol to expose tools (callable actions), resources (readable data), and prompts (reusable templates) to MCP-aware AI clients. Servers can run locally (stdio transport) or remotely (HTTP/SSE transport).</p>\n</blockquote>\n\n<h2>Two Server Transports</h2>\n<table>\n<thead><tr><th>Transport</th><th>Use Case</th><th>Example</th></tr></thead>\n<tbody>\n<tr><td><strong>stdio (local)</strong></td><td>The server runs on your machine; the AI client launches it as a subprocess</td><td>Filesystem MCP, local Postgres</td></tr>\n<tr><td><strong>HTTP / SSE (remote)</strong></td><td>The server runs on a vendor's infrastructure; the AI client connects over HTTPS</td><td>unLocked CRM MCP, GitHub MCP</td></tr>\n</tbody>\n</table>\n\n<h2>What an MCP Server Does (Lifecycle)</h2>\n<ol>\n<li><strong>Initialize</strong> — handshake with the AI client, exchange capabilities</li>\n<li><strong>List tools</strong> — return a JSON manifest of every callable action</li>\n<li><strong>Handle tool calls</strong> — execute the action and return structured results</li>\n<li><strong>Stream responses</strong> — for long-running tasks, push partial results via SSE</li>\n<li><strong>Manage auth</strong> — validate bearer tokens / OAuth scopes per tool</li>\n</ol>\n\n<h2>What a Tool Definition Looks Like</h2>\n<pre><code>{\n  \"name\": \"search_contacts\",\n  \"description\": \"Search insurance contacts by name, phone, email, or pipeline stage.\",\n  \"inputSchema\": {\n    \"type\": \"object\",\n    \"properties\": {\n      \"query\": { \"type\": \"string\" },\n      \"stage\": { \"type\": \"string\", \"enum\": [\"new\", \"qualified\", \"quoted\", \"bound\"] }\n    },\n    \"required\": [\"query\"]\n  }\n}</code></pre>\n\n<h2>Examples of MCP Servers in Production</h2>\n<ul>\n<li><strong>unLocked CRM MCP</strong> — 16 insurance tools (contacts, quotes, policies, commissions, SOAs)</li>\n<li><strong>GitHub MCP</strong> — read repos, create issues, manage PRs</li>\n<li><strong>Linear MCP</strong> — create and triage issues</li>\n<li><strong>Stripe MCP</strong> — read payments, refund charges</li>\n<li><strong>Filesystem MCP</strong> — read/write local files (local stdio)</li>\n</ul>\n\n<h2>Security Model</h2>\n<p>Remote MCP servers should:</p>\n<ul>\n<li>Authenticate every request (bearer token or OAuth 2.1)</li>\n<li>Scope tokens to a tenant / workspace</li>\n<li>Sign webhooks (HMAC-SHA256 with replay protection)</li>\n<li>Rate-limit per token</li>\n<li>Log every tool call for audit</li>\n</ul>\n\n<h2>How to Connect an MCP Server</h2>\n<p>For Claude Desktop, edit <code>~/Library/Application Support/Claude/claude_desktop_config.json</code>:</p>\n<pre><code>{\n  \"mcpServers\": {\n    \"unlocked-crm\": {\n      \"url\": \"https://mcp.unlockedcrm.ai/v1/YOUR_TENANT\",\n      \"headers\": { \"Authorization\": \"Bearer YOUR_TOKEN\" }\n    }\n  }\n}</code></pre>\n\n<h2>MCP Server vs MCP Client</h2>\n<p><strong>Server</strong> = the thing that exposes tools (your CRM). <strong>Client</strong> = the thing that calls them (Claude, ChatGPT, Cursor). One server can talk to many clients; one client can connect to many servers.</p>\n\n\n<h2>See unLocked CRM in Action</h2>\n<p>unLocked CRM ships a native MCP server, OpenAPI 3.1 spec, free A2P 10DLC registration, and remote SOA signing — built for the modern AI-first insurance agency. Book a 15-minute demo.</p>","category":"Definitions & Glossary","featureLink":"/mcp-server","cluster":"definitions","date":"2026-11-18","lastModified":"2026-11-18","readTime":"9 min read","author":"unLocked Team","tldr":"An MCP server is a program that exposes tools, resources, and prompts to AI models via the Model Context Protocol. Two transports: stdio (local) and HTTP/SSE (remote). Auth via bearer tokens / OAuth, tenant-scoped, with HMAC-signed webhooks.","definition":{"term":"MCP Server","text":"A program that implements the Model Context Protocol to expose tools, resources, and prompts to MCP-aware AI clients. Runs locally over stdio or remotely over HTTP/SSE."},"keyTakeaways":["Server exposes; client consumes","Two transports: stdio (local subprocess) and HTTP/SSE (remote)","Tools have JSON-Schema input definitions for safe calling","Production servers must scope tokens, sign webhooks, and log every call"],"faqs":[{"q":"Do I need to write my own MCP server?","a":"Only if you're a tool provider. Most users just connect existing MCP servers (unLocked CRM, GitHub, Linear) to their AI client."},{"q":"What language are MCP servers written in?","a":"Any language. Anthropic publishes official SDKs in TypeScript, Python, Java, C#, and Kotlin."},{"q":"Can one MCP server serve multiple users?","a":"Yes — remote MCP servers are typically multi-tenant, with each tenant identified via a scoped URL or bearer token."}]}