MCP
Model Context Protocol
The open protocol that lets AI agents use tools. Created by Anthropic, now governed by the Agentic AI Foundation (AAIF) under the Linux Foundation. The "USB-C of AI" — one standard, every tool.
What is MCP?
MCP (Model Context Protocol) is the standard way AI agents connect to external tools and data sources. Before MCP, every AI platform had to build custom integrations for every tool — a CRM connector for one platform wouldn't work on another. MCP solves this by providing a universal interface.
Think of it like USB: before USB, every device had its own connector. After USB, one connector works everywhere. MCP is the USB for AI tools.
How it works
An MCP server exposes tools (functions the agent can call), resources (data the agent can read), and prompts (templates the agent can use). The agent connects to the MCP server and discovers what's available — no manual configuration needed.
Agent → MCP Client → MCP Server → External Service
↓
Tools: search_contacts()
Resources: /crm/pipeline
Prompts: "Follow-up email"
Who created it?
Anthropic created MCP and open-sourced it under the Apache 2.0 license. In December 2025, Anthropic donated MCP to the Agentic AI Foundation (AAIF) under the Linux Foundation, co-founded with OpenAI and Block. It's now an industry standard, not a vendor-specific protocol.
Who supports it?
Anthropic (Claude), OpenAI (ChatGPT), Google (Gemini), Microsoft (Copilot), Cloudflare, Block, AWS, and hundreds of tool providers. As of 2026, MCP is the dominant protocol for agent-to-tool communication.
How MeetLoyd implements MCP
MeetLoyd has 100% MCP coverage — every agent-to-tool interaction goes through MCP. This means:
- OAuth + SSO integrations — Google Workspace, Microsoft 365, HubSpot, Salesforce, GitHub, Slack, and more. Each exposed as an MCP server.
- Custom MCP servers — Build your own tools via the Blueprint Studio or bring existing MCP servers.
- Authorization layer — Every MCP tool call passes through MeetLoyd's 106-permission RBAC. Default deny. Fail closed.
- Audit trail — Every tool invocation is logged: who called what, with what parameters, what the result was, and what it cost.
Unlike raw MCP (which has no auth, no governance, no audit), MeetLoyd wraps MCP with enterprise controls. You get the interoperability of an open standard with the security of enterprise governance.
MCP vs other protocols
MCP is specifically for agent-to-tool communication. It's complementary to:
- A2A — Agent-to-Agent communication (how agents talk to each other)
- SLIM — Cross-organization agent messaging (how agents from different companies collaborate)
- UCP — Agent commerce (how agents buy and sell)
MeetLoyd implements all four. Together, they form the protocol stack for the Internet of Agents.
Key concepts
Tools
Functions the agent can execute. Example: search_contacts(query: "John") calls the CRM and returns matching contacts. Tools have typed parameters and return structured results.
Resources
Data the agent can read. Example: crm://pipeline/active-deals returns current deal data. Resources are read-only and described by URI templates.
Prompts
Reusable templates that guide agent behavior. Example: a "follow-up email" prompt with variables for the contact name and last interaction.
Transport
MCP uses Server-Sent Events (SSE) for streaming and HTTP for stateless communication. The latest spec (Nov 2025) added support for async operations, statelessness, and server identity.