[GTM glossary](/glossary.txt) MCP server Contextual Intelligence & AI # MCP server (MCP) In short An MCP server is a lightweight service that makes a tool or data source available to AI applications using the Model Context Protocol. It describes what it can do, and an AI assistant can then call those functions on a user's behalf. The protocol is the standard; the server is one concrete implementation for one system. A team might run one server for its CRM, another for a company database, and another for its calendar, and connect an assistant to all of them. ## What an MCP server provides * **Tools:** Actions the assistant can call, such as search accounts, create a task, or look up a contact. * **Resources:** Read-only data the assistant can load, such as a document or a record. * **Prompts:** Reusable instruction templates for common jobs. --- ## MCP server vs. MCP client The client lives inside the AI application and asks for capabilities. The server wraps your system and answers. The server decides what is exposed and under what permissions. --- ## Example in a sales workflow Say a rep asks an assistant, "Which of my open accounts hired a new VP of Sales this month?" The assistant calls a data server to find matches, calls a CRM server to filter to the rep's accounts, and returns a short list. No one exported a spreadsheet or switched tabs. --- ## What to check before you connect one * **Permissions:** Prefer read-only access first. Grant write actions only where a mistake is easy to undo. * **Authentication:** Confirm it uses your existing login and respects each user's own access rights. * **Data handling:** Know what the server sends to the model provider and whether it is logged. * **Source trust:** Only connect servers you built or have reviewed. A malicious server or poisoned data can steer an assistant. --- ## Common mistakes * **Exposing too much:** Giving an assistant full database access instead of narrow tools. * **No audit trail:** Skipping logs of which actions the assistant took. * **Too many tools:** Long tool lists confuse the model and reduce accuracy. --- ## MCP server FAQs **What is an MCP server?** An MCP server is a program that exposes a system's data and actions to AI assistants using the Model Context Protocol. It lists what it can do, such as searching records or creating tasks, so an assistant can use those functions on request. **What does an MCP server do for sales teams?** It lets an AI assistant query and act on tools sales teams already use, such as a CRM or company database, through natural language. Reps can ask questions and trigger tasks without exporting data or switching between applications. **Is an MCP server safe to use?** Safety depends on setup. Use servers you trust, start with read-only access, rely on your existing authentication, and keep logs of actions. Review what data is sent to the model provider before connecting sensitive systems. **How is an MCP server different from an API?** An API is a general interface a developer codes against. An MCP server wraps tools in a standard format that AI assistants can discover and call without custom integration work. Many MCP servers use an API underneath. Share term --- ## More related phrases * [CRM integration](/glossary/crm-integration.txt) * [Custom API integration](/glossary/custom-api-integration.txt) * [Native integration](/glossary/native-integration.txt) ## Similar phrases * [Model Context Protocol (MCP)](/glossary/model-context-protocol.txt) The Model Context Protocol (MCP) is an open standard that lets AI assistants connect to external tools and data sources in a consistent way, instead of needing a custom integration for each one. * [Agentic workflow](/glossary/agentic-workflow.txt) An agentic workflow is a multi-step process in which an AI agent decides the next action, uses tools, and checks its work, often with human approval at set checkpoints. * [AI agent](/glossary/ai-agent.txt) An AI agent is software powered by an AI model that takes a goal, uses tools and data to act on it, and decides its own next steps. * [Context layer](/glossary/context-layer.txt) A context layer is the shared data and meaning layer between your systems and your AI tools, giving them consistent definitions, relationships, and fresh facts to work from. * [Multi-agent system](/glossary/multi-agent-system.txt) A multi-agent system uses several specialized AI agents that work together, passing tasks and results between them, to complete work that would be too broad for a single agent.