MCP Server Quick Start

Note: The MCP Server functionality described in this document requires Higress 2.1.0 or higher.

Higress AI Gateway provides unified MCP Server hosting capabilities, helping AI Agents quickly integrate with various data sources. Through MCP Server, AI Agents can easily access databases, REST APIs, and other external services without concerning themselves with specific connection details. Database integration is a built-in capability of the gateway, while any external REST API can be transformed into an MCP Server through simple configuration. This guide will demonstrate the configuration process using PostgreSQL database and a simple REST API as examples.

Prerequisites

Before using MCP Server, you need to deploy Higress first. You can refer to the Higress Quick Start Guide for deployment instructions. When installing Higress, make sure to add the following parameter to your Helm command:

helm install higress -n higress-system [...other parameters...] --set global.enableRedis=true

MCP Server requires Redis service for data caching. After enabling it, you can check the Redis service address using the following command:

kubectl get svc redis-stack-server -n higress-system -o wide

Configuring MCP Server

ConfigMap Global Parameter Configuration

Configure MCP Server-related global parameters in the ConfigMap:

kubectl edit configmap higress-config -n higress-system

Configure Redis connection information and MCP Server routing rules:

apiVersion: v1
data:
  higress: |-
    mcpServer:
      sse_path_suffix: /sse  # Path suffix for SSE connections
      enable: true          # Enable MCP Server
      redis:
        address: redis-stack-server.higress-system.svc.cluster.local:6379 # Redis service address
        username: "" # Redis username (optional)
        password: "" # Redis password (optional)
        db: 0 # Redis database (optional)
      match_list:          # MCP Server Session Persistence Routing Rules (When the following paths are matched, they will be recognized as an MCP session, and session persistence will be maintained through mechanisms such as SSE)
        - match_rule_domain: "*"
          match_rule_path: /postgres
          match_rule_type: "prefix"
        - match_rule_domain: "*"
          match_rule_path: /user
          match_rule_type: "prefix"
      servers: []
...

kind: ConfigMap
metadata:
  name: higress-config
  namespace: higress-system

Note: Database-type MCP Servers are configured in the Config Map, while REST API type configurations are done in the Higress console.

Configuring Database MCP Server

Configure Database MCP Server in the Config Map:

servers:
  - name: postgres      # MCP Server name
    path: /postgres     # Access path, must match the configuration in match_list
    type: database      # Type is database
    config:
      dsn: "your postgres database connect dsn" # Database connection string
      dbType: "postgres"                        # Database types, currently supported: postgres/mysql/clickhouse/sqlite

For database connection string format, please refer to the gorm documentation.

Configuring REST API MCP Server

Any REST API can be quickly transformed into an MCP Server through the following steps:

1. Add Service Source

Add the target REST API’s service source in the Higress console. This example uses randomuser.me as the service source:

2. Configure Route

Add a route in the Higress console and point it to the corresponding service source:

3. Configure MCP Server Plugin

Click the Strategy button in the action of the route.

Add and configure the MCP Server plugin for the created route:

Plugin configuration example:

server:
  name: "random-user-server"
tools:
- description: "Get random user information"
  name: "get-user"
  requestTemplate:
    method: "GET"
    url: "https://randomuser.me/api/"
  responseTemplate:
    body: |-
      # User Information
      {{- with (index .results 0) }}
      - **Name**: {{.name.first}} {{.name.last}}
      - **Email**: {{.email}}
      - **Location**: {{.location.city}}, {{.location.country}}
      - **Phone**: {{.phone}}
      {{- end }}

For more detailed information about configuring REST API to MCP Server, please refer to the MCP Server Plugin Configuration Reference.

Note: For the 2025-03-26 MCP streamable HTTP protocol, ConfigMap configuration is not required.

Using MCP Server

Configure the MCP Server SSE connection in your AI Agent. Taking cursor as an example:

• For database-type MCP Server: use the path + sse_path_suffix configured in ConfigMap
• For REST API type MCP Server: use the route path + sse_path_suffix configured in the console

"mcpServers": {
  "postgres": {
    "url": "http://your-higress-address/postgres/sse"
  },
  "rest-api": {
    "url": "http://your-higress-address/user/sse"
  }
}

Cursor configuration complete:

Through MCP Server, you can quickly add various data source support for AI Agents, improving development efficiency. Any REST API can be transformed into an MCP Server through simple configuration, without writing additional code.

If you encounter any issues while using MCP Server, please leave your information in Higress Github Issue.

If you are interested in future updates of Higress or would like to provide feedback, welcome to Star Higress Github Repo.