A Python-based Model Context Protocol (MCP) server and companion client for configuring and managing Higress through LangGraph/LangChain tooling.
https://github.com/higress-group/higress-ops-mcp-serverSkip the kubectl gymnastics and API documentation deep-dives. This MCP server connects your AI assistant directly to Higress, turning complex ingress operations into conversational commands with built-in safety rails.
Managing Higress typically means juggling multiple interfaces: the console, kubectl commands, API references, and YAML configurations. You know the drill—checking route configurations in one tab, debugging service sources in another, then switching to update plugins somewhere else.
This MCP server consolidates all of that into a single conversational interface. Tell your AI assistant to "check the routes for the user-service" or "add rate limiting to the checkout API," and it handles the underlying Higress API calls automatically.
Route Management: Query, create, and update routes without diving into YAML. Ask your assistant to show you all routes pointing to a specific service, or quickly add a new route with specific matching rules.
Service Discovery: Instantly check service source configurations, add new upstream services, or troubleshoot connectivity issues—all through natural language queries.
Request Filtering: Configure request blocking policies conversationally. Instead of hunting through plugin documentation, just describe what you want to block and let the AI handle the configuration details.
Safety-First Operations: Sensitive operations like adding routes or updating plugins require human confirmation before execution. You get the speed of AI assistance with the safety of manual approval for critical changes.
Incident Response: During an outage, ask your assistant to quickly check which routes are failing and block problematic traffic patterns while you investigate—no need to remember specific API endpoints or navigate multiple dashboards.
Feature Rollouts: When deploying new services, describe your routing requirements in plain English: "Route all /api/v2 requests to the new microservice, but keep /api/v1 on the old one." The assistant handles the Higress configuration complexity.
Daily Operations: Check service health, review current route configurations, or audit request blocking rules without switching between tools. Your assistant becomes your unified Higress command center.
Built on LangGraph and LangChain MCP adapters, this isn't just another API wrapper. The client includes intelligent agent flows that understand context and can chain operations together. The server runs locally or in your infrastructure, so your Higress credentials never leave your environment.
Setting up takes minutes: copy the environment config, point it at your Higress instance, and run uv run client.py. The MCP server handles the protocol complexity while you focus on managing your infrastructure.
Need custom operations? The architecture makes adding new tools straightforward. Create a tool class, implement your Higress API calls, register it with the server, and you're done. The documentation includes clear examples for extending functionality—no need to understand MCP internals.
Whether you're managing a handful of services or orchestrating complex microservice routing, this MCP server transforms Higress from a configuration-heavy platform into a conversational partner for your infrastructure operations.