TocharianOU/mcp-server-kibana

Kibana MCP Server

API Specification

This project is based on the official Elastic Kibana API documentation and uses the OpenAPI YAML specification from Elastic Stack 8.x (ES8) to dynamically retrieve and manage all Kibana API endpoints. For the latest details, see the Kibana API documentation.

A Kibana MCP server implementation that allows any MCP-compatible client (such as Claude Desktop) to access your Kibana instance via natural language or programmatic requests.

This project is community-maintained and is not an official product of Elastic or MCP.

🚀 Installation

Quick Install

# Global installation (recommended)
npm install -g @tocharian/mcp-server-kibana

# Or local installation
npm install @tocharian/mcp-server-kibana

Alternative: From Source

git clone https://github.com/TocharianOU/mcp-server-kibana.git
cd mcp-server-kibana
npm install
npm run build

🎯 Quick Start

Method 1: Direct CLI Usage

Using Basic Authentication

# Set your Kibana credentials and run
KIBANA_URL=http://your-kibana-server:5601 \
KIBANA_USERNAME=your-username \
KIBANA_PASSWORD=your-password \
npx @tocharian/mcp-server-kibana

Using Cookie Authentication

# Set your Kibana session cookies and run
KIBANA_URL=http://your-kibana-server:5601 \
KIBANA_COOKIES="sid=your-session-id; security-session=your-security-session" \
npx @tocharian/mcp-server-kibana

Method 2: Claude Desktop Integration (Recommended)

Add to your Claude Desktop configuration file:

Config file locations:

• MacOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json

Using Basic Authentication

{
  "mcpServers": {
    "kibana-mcp-server": {
      "command": "npx",
      "args": ["@tocharian/mcp-server-kibana"],
      "env": {
        "KIBANA_URL": "http://your-kibana-server:5601",
        "KIBANA_USERNAME": "your-username",
        "KIBANA_PASSWORD": "your-password",
        "KIBANA_DEFAULT_SPACE": "default",
        "NODE_TLS_REJECT_UNAUTHORIZED": "0"
      }
    }
  }
}

Using Cookie Authentication

{
  "mcpServers": {
    "kibana-mcp-server": {
      "command": "npx",
      "args": ["@tocharian/mcp-server-kibana"],
      "env": {
        "KIBANA_URL": "http://your-kibana-server:5601",
        "KIBANA_COOKIES": "sid=your-session-id; security-session=your-security-session",
        "KIBANA_DEFAULT_SPACE": "default",
        "NODE_TLS_REJECT_UNAUTHORIZED": "0"
      }
    }
  }
}

Method 3: Using Environment File

# Create .env file
cat > kibana-mcp.env << EOF
KIBANA_URL=http://your-kibana-server:5601
KIBANA_USERNAME=your-username
KIBANA_PASSWORD=your-password
NODE_TLS_REJECT_UNAUTHORIZED=0
EOF

# Run with environment file
env $(cat kibana-mcp.env | xargs) npx @tocharian/mcp-server-kibana

Method 4: Streamable HTTP Mode (NEW in v0.4.0)

Run the server as a standalone HTTP service for remote access and API integration:

# Start HTTP server (default port 3000)
MCP_TRANSPORT=http \
KIBANA_URL=http://your-kibana-server:5601 \
KIBANA_USERNAME=your-username \
KIBANA_PASSWORD=your-password \
npx @tocharian/mcp-server-kibana

# Or with custom port and host
MCP_TRANSPORT=http \
MCP_HTTP_PORT=9000 \
MCP_HTTP_HOST=0.0.0.0 \
KIBANA_URL=http://your-kibana-server:5601 \
KIBANA_USERNAME=your-username \
KIBANA_PASSWORD=your-password \
npx @tocharian/mcp-server-kibana

HTTP Mode Features:

• Exposes MCP server at http://host:port/mcp endpoint
• Health check available at http://host:port/health
• Session-based connection management
• Supports both POST (JSON-RPC requests) and GET (SSE streams)
• Compatible with any HTTP client or MCP SDK

Example HTTP client usage:

// Initialize connection
const response = await fetch('http://localhost:3000/mcp', {
  method: 'POST',
  headers: { 'Content-Type': 'application/json' },
  body: JSON.stringify({
    jsonrpc: '2.0',
    method: 'initialize',
    params: {
      protocolVersion: '2024-11-05',
      capabilities: {},
      clientInfo: { name: 'my-client', version: '1.0.0' }
    },
    id: 1
  })
});

const sessionId = response.headers.get('mcp-session-id');

// Subsequent requests include session ID
const toolsResponse = await fetch('http://localhost:3000/mcp', {
  method: 'POST',
  headers: {
    'Content-Type': 'application/json',
    'mcp-session-id': sessionId
  },
  body: JSON.stringify({
    jsonrpc: '2.0',
    method: 'tools/list',
    params: {},
    id: 2
  })
});

Features

Core Features

• Connect to local or remote Kibana instances
• Dual transport modes:
  • Stdio transport (default) - For Claude Desktop and local MCP clients
  • Streamable HTTP transport (NEW in v0.4.0) - For remote access, API integration, and web applications
• Dual authentication support:
  • Cookie-based authentication (recommended for browser sessions)
  • Basic authentication (username/password)
• SSL/TLS and custom CA certificate support
• Multi-space support for enterprise Kibana environments
• Exposes Kibana API endpoints as both tools and resources
• Search, view, and execute Kibana APIs from MCP clients
• Type-safe, extensible, and easy to integrate
• Session management with automatic UUID generation for HTTP mode
• Health check endpoint for monitoring and load balancing

Visualization Layer (VL) Features

• Complete CRUD operations for Kibana saved objects
• Universal saved object management - works with all object types
• Intelligent parameter handling - supports multiple input formats (arrays, JSON strings, comma-separated)
• Optimized search with pagination support and performance tips
• Bulk operations for efficient mass updates and deletions
• Version control with optimistic concurrency for safe updates
• Reference management for object relationships
• Multi-format type support - flexible input parsing for better UX

Directory Structure

├── index.ts                # Server entry point
├── src/
│   ├── types.ts            # Type definitions and schemas
│   ├── base-tools.ts       # Tool registration and API logic
│   ├── prompts.ts          # Prompt registration (expert & resource helper)
│   ├── resources.ts        # Resource registration (API paths/URIs)
│   ├── vl_search_tools.ts  # Visualization Layer - Search tools
│   ├── vl_get_tools.ts     # Visualization Layer - Get tools
│   ├── vl_create_tools.ts  # Visualization Layer - Create tools
│   ├── vl_update_tools.ts  # Visualization Layer - Update tools
│   └── vl_delete_tools.ts  # Visualization Layer - Delete tools
├── kibana-openapi-source.yaml # Kibana API OpenAPI index
├── README.md               # English documentation
├── README_zh.md            # Chinese documentation

Resources

Examples:

• kibana-api://paths?search=saved_objects
• kibana-api://path/GET/%2Fapi%2Fstatus

Tools

Base Tools

Visualization Layer (VL) Tools - Saved Objects Management

Supported Saved Object Types: dashboard, visualization, index-pattern, search, config, lens, map, tag, canvas-workpad, canvas-element

Prompts

Configuration

Configure the server via environment variables:

Kibana Connection Settings

*Either KIBANA_COOKIES or both KIBANA_USERNAME and KIBANA_PASSWORD must be provided for authentication.

Transport Mode Settings (NEW in v0.4.0)

Transport Mode Details:

• Stdio mode (default): For Claude Desktop and local MCP clients
• HTTP mode: Runs as a standalone HTTP server for remote access, API integration, and web applications

📦 Package Information

• NPM Package: @tocharian/mcp-server-kibana
• GitHub Repository: TocharianOU/mcp-server-kibana
• Node.js: >= 18.0.0
• Package Size: ~685KB (6.4MB unpacked)

🔧 Troubleshooting

Common Issues

"import: command not found" error

# Make sure you're using the latest version
npm install -g @tocharian/mcp-server-kibana@latest

# Or try using node directly
node $(which mcp-server-kibana)

Connection issues

• Verify Kibana URL is accessible
• Check authentication credentials
• For SSL issues, try setting NODE_TLS_REJECT_UNAUTHORIZED=0

Claude Desktop not detecting the server

• Restart Claude Desktop after config changes
• Check config file syntax with a JSON validator
• Verify environment variables are set correctly

Example Queries

Basic Queries

• "What is the status of my Kibana server?"
• "What is the status of my Kibana server in the 'marketing' space?"
• "List all available Kibana spaces I can access."
• "List all available Kibana API endpoints."
• "Show details for the POST /api/saved_objects/_find endpoint."
• "Execute a custom API request for /api/status."

Saved Objects Management

• "Search for all dashboards in Kibana"
• "Find visualizations containing 'nginx' in the title"
• "Get dashboard with ID 'my-dashboard-123'"
• "Create a new dashboard with title 'Sales Overview'"
• "Update the description of visualization 'viz-456'"
• "Delete multiple old dashboards by their IDs"
• "Search for lens visualizations in the 'analytics' space"
• "Find all canvas workpads created this month"
• "Get the first 10 index patterns with only title and description fields"
• "Bulk update multiple dashboard titles"
• "Search across multiple object types: dashboards and visualizations"
• "Create a new index pattern for 'logs-*' with timestamp field"

Two Prompt Modes in Claude Desktop

When using this server with Claude Desktop, two different prompt interaction modes are supported:

1. Tool-based Prompt Mode

• How it works: Claude Desktop can directly call server tools (including base tools like get_status, execute_api, and VL tools like vl_search_saved_objects, vl_create_saved_object) to answer your questions or perform actions.
• Best for: Users who want a conversational, guided experience. The server will automatically search, execute, and explain Kibana APIs and manage saved objects.
• Example: "Show all dashboards containing 'sales'" or "Create a new visualization for web analytics"
• Testing tip: Select the kibana-tool-expert prompt in Claude Desktop for integration testing, then start using it.

2. Resource-based Prompt Mode

• How it works: Claude Desktop interacts with the server via resource URIs (such as kibana-api://paths or kibana-api://path/GET/%2Fapi%2Fstatus), and the server returns structured data for Claude to parse.
• Best for: Advanced users, MCP clients that only support resource access, or programming scenarios needing raw API metadata.
• Example: "Get resource kibana-api://paths?search=dashboard"

Note: The two endpoints in resources (kibana-api://paths and kibana-api://path/{method}/{encoded_path}) have corresponding base tools (list_all_kibana_api_paths, get_kibana_api_detail). This design ensures compatibility with MCP clients that cannot intelligently select multiple resources, making it easier for tools like Claude Desktop to interact with Kibana.

Tip: Most users are recommended to use tool mode for a more natural and powerful experience; resource mode offers maximum flexibility for advanced and compatibility use cases.

Development

Install dependencies:

npm install

Build the server:

npm run build

Auto-rebuild in development mode:

npm run watch

Run in different modes:

# Stdio mode (default)
npm start

# HTTP mode
npm run start:http

# Development with TypeScript
npm run start:ts

# HTTP mode with TypeScript
npm run start:http:ts

Debugging

Since the MCP server communicates via stdio, debugging can be inconvenient. It is recommended to use MCP Inspector:

npm run inspector

After starting, Inspector will provide a browser-accessible debugging tool URL.

Community

This project is community-maintained. Contributions and feedback are welcome! Please be respectful and inclusive in all communications, and follow the Elastic Community Code of Conduct.

License

This project is licensed under the Apache License 2.0. See the file for details.

Troubleshooting

• Check if MCP configuration is correct
• Ensure the Kibana address is accessible
• Verify authentication credentials have sufficient permissions
• If using a custom CA, ensure the certificate path is correct and readable
• If using NODE_TLS_REJECT_UNAUTHORIZED=0, be aware of security risks
• Check error messages output in the terminal