Kong Konnect MCP Server

The Kong Konnect MCP Server enables developers to manage API infrastructure and debug performance issues directly from their development environment. You can also plug in Kong Konnect to your AI agents to automate tasks like configuration management and performance troubleshooting.

The same tools available in Kong Konnect MCP Server power KAi (Kong’s AI assistant), an in-product AI assistant for Kong Konnect.

Kong Konnect MCP server is in active development. Expect continuous updates and new tools to be added regularly.

Use cases

The Kong Konnect MCP Server enables several workflows for managing and debugging your API infrastructure:

• Kong Konnect entity management: Query control planes, services, routes, consumers, consumer groups, plugins, and vaults.
• API debugging: Create debug sessions with active tracing to investigate performance issues and identify bottlenecks.
• Analytics and monitoring: Query API request data with filters for time range, status codes, consumers, services, and routes.
• Kong documentation search: Search Kong’s documentation for configuration guidance and troubleshooting steps.

When should I use an MCP Server?

Use the Kong Konnect MCP Server when:

• You want to manage Konnect resources from your IDE or terminal
• You need to debug API performance issues with active tracing
• You prefer working with AI assistants for infrastructure tasks
• You need to query analytics and traffic patterns programmatically

Kong Konnect MCP Server provides programmatic access to your Konnect resources through the Model Context Protocol. The server exposes specialized tools that AI assistants can use to query Kong Gateway entities, analyze traffic, debug performance issues, and search documentation.

You can connect to the MCP server from:

• IDEs with MCP client support (VS Code, Cursor, IntelliJ)
• AI assistants like Claude and GitHub Copilot
• Claude Code CLI for terminal-based workflows
• Any MCP-compatible client application

The server connects to Konnect backend services to retrieve Kong Gateway configuration, analytics data, active tracing sessions, and Kong’s documentation. This allows AI assistants to provide contextual assistance for debugging, configuration, and operational tasks.

PAT-based authentication

The MCP server can authenticate using your Personal Access Token (PAT), which means it only accesses resources you have permission to view. Create a token by opening the Kong Konnect tokens page and selecting Generate Token. All queries respect your organization’s role-based access controls and regional data boundaries.

SPAT-based authentication

For automated workflows, you can also use a Service Personal Access Token (SPAT). SPATs are tied to specific service accounts with defined permissions, allowing secure machine-to-machine authentication without user intervention.

The MCP server is deployed regionally. Connect to the server in the same region where your Konnect resources are deployed.

Default endpoint

The regional endpoint defaults to US region.

Organizations using multiple Kong Konnect regions require separate MCP server connections for each region. Resources cannot be accessed across regions from a single connection.

Organization settings

Kong Konnect MCP Server access is enabled by default. Organization administrators can disable it from the Organization Settings.

Configure the MCP client of your choice by adding the Kong Konnect MCP Server with your regional URL and PAT. Select your preferred client below for specific installation instructions.

The Kong Konnect MCP Server provides tools for managing Kong Gateway entities, debugging performance, and accessing documentation. Tool availability depends on your user permissions and organization entitlements.

View detailed tool reference

Pattern 1: Debugging performance issues

Create a debug session and analyze trace data to identify bottlenecks.

1. GetControlPlane (operation="list" or "get_by_name")
   → Select target control plane
   → Extract control_plane_id

2. CreateDebugSession (
     control_plane_id=<from_step_1>,
     http_path="/slow-endpoint",
     http_latency=">=1000ms",
     session_duration=60
   )
   → Confirm with user
   → Extract debug_session_id

3. ActiveTracingSession (
     control_plane_id=<from_step_1>,
     debug_session_id=<from_step_2>,
     operation="fetch_status"
   )
   → Wait for status="completed"

4. ActiveTracingSession (
     control_plane_id=<from_step_1>,
     debug_session_id=<from_step_2>,
     operation="summarize_session"
   )
   → Analyze bottlenecks and latency distribution

Pattern 2: Investigating an API endpoint

Trace an API path through control plane, route, service, and plugin configuration.

1. GetControlPlane (operation="get_by_route", path="/api/users")
   → Extract control_plane_id

2. GetRoute (control_plane_id=<from_step_1>, operation="list")
   → Find route matching the path
   → Extract service_id from route

3. GetService (control_plane_id=<from_step_1>, operation="get_by_id", service_id=<from_step_2>)
   → Review upstream configuration

4. GetPlugin (control_plane_id=<from_step_1>, operation="list")
   → Check plugins affecting the route/service

Pattern 3: Consumer-specific analysis

Identify failing consumers and analyze their request patterns.

1. GetControlPlane (operation="list")
   → Select target control plane
   → Extract control_plane_id

2. GetAnalytics (
     time_range="24H",
     operation="query_api_requests",
     status_codes=[500, 502, 503]
   )
   → Identify top failing consumers from response

3. GetConsumer (
     control_plane_id=<from_step_1>,
     operation="get_by_id",
     consumer_id=<from_step_2_response>
   )
   → Review consumer configuration

4. GetAnalytics (
     time_range="24H",
     operation="get_consumer_requests",
     consumer_id=<from_step_3>,
     failureOnly=true
   )
   → Analyze consumer-specific failures

5. GetPlugin (
     control_plane_id=<from_step_1>,
     operation="list"
   )
   → Filter for consumer-scoped auth plugins

Pattern 4: Configuration audit

Audit Kong Gateway configuration across all control planes.

1. GetControlPlane (operation="list")
   → Iterate through all control planes

2. For each control plane:
   a. GetService (control_plane_id=<current>, operation="list")
   b. GetRoute (control_plane_id=<current>, operation="list")
   c. GetPlugin (control_plane_id=<current>, operation="list")
   d. GetConsumer (control_plane_id=<current>, operation="list")
   e. GetVault (control_plane_id=<current>, operation="list")

3. Aggregate and analyze configurations across all control planes