MCP Module

Introduction

MCP (Model Context Protocol) is a special interface provided by Nginx UI that allows AI agents to interact with Nginx UI. Through MCP, AI models can access and manage Nginx configuration files, perform Nginx-related operations (such as restart, reload), and get Nginx running status.

Feature Overview

The MCP module is divided into two main functional areas:

• Configuration File Management - Various operations for managing Nginx configuration files
• Nginx Service Management - Control and monitor Nginx service status

Interface

The MCP interface is accessible through /mcp using Streamable HTTP. Configure clients such as OpenCode with this URL and send JSON-RPC requests using HTTP POST. Initialization returns a JSON-RPC response directly, without first opening an SSE connection.

For protocol versions that use sessions, initialization returns a Mcp-Session-Id response header. Include this header on subsequent requests, including GET requests that open a server event stream. Clients can also identify Streamable HTTP GET requests using the MCP-Protocol-Version header.

Legacy HTTP+SSE clients remain supported: a GET to /mcp without either header opens the legacy SSE stream, which announces /mcp_message?sessionId=... as its message endpoint. Both transports use the same authentication and tool scopes.

Authentication

Create a service token in Preferences > Access Tokens and grant it the smallest MCP scope the client needs:

• mcp:read permits resources and read-only tools.
• mcp:write permits mutating tools and includes MCP read access.

Send the token in the Authorization header:

Authorization: Bearer nui_pat_...

Tokens are displayed only once. Store them in a secret manager and set an expiration date whenever possible. Credentials in URL query parameters are rejected to prevent leakage through logs and browser history.

Resources

Resources are readable information provided by MCP, such as Nginx status.

Tools

Tools are executable operations provided by MCP, such as restarting Nginx, modifying configuration files, etc.

Use Cases

MCP is mainly used in the following scenarios:

1. AI-driven Nginx configuration management
2. Integration with automated operations tools
3. Integration of third-party systems with Nginx UI
4. Providing machine-readable APIs for automation scripts