Ep 21: The Infinite Toolbox — MCP Protocol & n8n Integration Architecture
What Problem Does MCP Solve?
Before MCP, every AI app needed custom integrations for every external service (N×M problem). MCP standardizes this into N+M.
graph TB
subgraph "After MCP: N + M standardized"
B1[Claude] --> MCP["🔌 MCP Protocol"]
B2[ChatGPT] --> MCP
B3[n8n Agent] --> MCP
MCP --> T1[Gmail MCP Server]
MCP --> T2[Slack MCP Server]
MCP --> T3[GitHub MCP Server]
end
style MCP fill:#22c55e,stroke:#16a34a,color:#fffMCP is to AI Agents what USB is to computers — a standardized interface protocol for plug-and-play connectivity.
1. MCP Architecture
graph TB
Client["🤖 MCP Client (n8n)"] -->|"Discover tools"| Protocol["📡 JSON-RPC 2.0"]
Protocol -->|"Call tools"| Server["🔧 MCP Server"]
Server --> Tools["🔧 Tools: executable actions"]
Server --> Resources["📂 Resources: readable data"]
Server --> Prompts["📝 Prompts: templates"]
style Protocol fill:#f59e0b,stroke:#d97706,color:#fff2. n8n's Dual Role
n8n is both MCP Client (consume external servers) and MCP Server (expose workflows as tools).
3. MCP vs REST APIs
| Dimension | REST API | MCP |
|---|---|---|
| Discovery | Read docs manually | Auto-negotiation |
| Parameters | Varies per API | JSON Schema standardized |
| AI-ready | Write Tool Descriptions | Self-describing, AI auto-understands |
Next Episode
Ep 22: hands-on connecting MCP Client Tool to GitHub, Filesystem and other external MCP servers.