What Are MCP Servers?

MCP (Model Context Protocol) is an open standard for connecting AI models to external tools and data sources. MCP servers act as bridges between your agents and the outside world.

When an agent needs to query Google Analytics, search Ahrefs, or call a custom API, it does so through an MCP server.

How It Works

Agent → MCP Server → External Service
                  ← Data/Results

The agent decides it needs external data (e.g., website traffic)
It calls an MCP tool (e.g., mcp__google-analytics__runReport)
The MCP server handles authentication and API calls
Results are returned to the agent

MCP servers run as separate processes alongside the agent. They're started when a conversation begins and stopped when it ends.

Types of MCP Servers

stdio (Standard I/O)

The most common type. The MCP server runs as a child process, communicating via stdin/stdout.

{
  "type": "stdio",
  "command": "npx",
  "args": ["-y", "@anthropic-ai/mcp-google-analytics"],
  "env": {
    "GA_PROPERTY_ID": "12345678"
  }
}

SSE (Server-Sent Events)

Remote MCP servers accessed over HTTP. Useful for shared infrastructure.

{
  "type": "sse",
  "url": "https://mcp.example.com/sse"
}

What MCP Servers Provide

Capability	Description	Example
Tools	Functions the agent can call	`runReport`, `searchKeywords`
Resources	Data the agent can read	Configuration files, schemas
Prompts	Pre-built prompt templates	Report templates

Most MCP servers primarily provide tools. When installed, these tools appear with the naming pattern:

mcp__{server-name}__{tool-name}

For example:

mcp__ahrefs__site-explorer-domain-rating
mcp__google-analytics__runReport
mcp__search-console__get_top_queries

Built-in Platform Tools

TeamDay provides three MCP servers automatically — no installation needed:

Server	Tool Name	Purpose
teamday-media	`MediaGeneration`	Generate images and videos
teamday-admin	`TeamdayAdmin`	Manage platform resources
teamday-ui	`UICommand`	Control the chat interface

See Platform Tools for details.

Popular MCP Integrations

Integration	What It Does
Google Analytics	Query GA4 traffic data, page performance, user behavior
Google Search Console	Search impressions, clicks, CTR, keyword rankings
Ahrefs	Backlinks, domain rating, keyword research, competitor analysis
Slack	Send messages, read channels, manage threads
GitHub	Issues, PRs, code search, repository management
PostgreSQL / MySQL	Database queries and schema inspection
Filesystem	Read/write files outside the sandbox

MCP vs Skills

	MCP Servers	Skills
Purpose	Connect to external services	Teach agents workflows
Loading	At conversation start	On-demand (progressive discovery)
Format	Running process (stdio/SSE)	Markdown file + optional scripts
Best for	API integrations, databases	Local automation, domain expertise

Rule of thumb: If it requires a network call to an external service, use an MCP. If it's a local workflow or set of instructions, use a Skill.

Next Steps

Installing MCP Servers — Set up MCP integrations on your Spaces
Creating Custom MCP Servers — Build your own MCP server
Tools Overview — All tool categories