elastic/example-mcp-dashbuilder

开发工具官方

本页由 AI 翻译，技术细节请以原文为准。查看原文

18 stars 6 forks 质量 76 趋势 30NOASSERTION

Repository Install

MCP app to build dashboards via cursor

概览

An MCP (Model Context Protocol) app that lets AI assistants build Kibana dashboards using ES|QL and Elastic Charts. Create visualizations through natural language, preview them live with Kibana's grid layout, and export directly to Kibana as real Lens dashboards. Screenshot from Claude Desktop: the MCP app renders the interactive dashboard inside the chat. - — describe what you want, the AI builds it - — open-ended exploration flow that runs aggregations, surfaces patterns, builds charts for key findings, and suggests drill-down queries (triggered by prompts like "analyze my X") - — all queries use ES|QL for data retrieval - — full interactive dashboard rendered directly in the chat via MCP Apps - — same 48-column drag-and-drop grid as Kibana dashboards - — matches Kibana's latest visual design - — organize panels into groups - — creates real Kibana dashboards with Lens visualizations - — import existing ES|QL-based Kibana dashboards for AI-assisted editing - — apply custom palettes...

README

以下为原文 README 节选，完整内容请查看原文。

example-mcp-dashbuilder

Screenshot from Claude Desktop: the MCP app renders the interactive dashboard inside the chat.

What it does

You (in an MCP client) → "Build me an ecommerce analytics dashboard"
    ↓
AI explores your Elasticsearch data via ES|QL
    ↓
Creates charts (bar, line, pie, metric, heatmap) with Elastic Charts
    ↓
Interactive dashboard rendered inline in the chat (MCP Apps)
    ↓
One-click export to Kibana as Lens visualizations

Features

Natural language dashboard creation — describe what you want, the AI builds it
Deep data analysis — open-ended exploration flow that runs aggregations, surfaces patterns, builds charts for key findings, and suggests drill-down queries (triggered by prompts like “analyze my X”)
ES|QL powered — all queries use ES|QL for data retrieval
Inline dashboard preview — full interactive dashboard rendered directly in the chat via MCP Apps
Kibana grid layout — same 48-column drag-and-drop grid as Kibana dashboards
Borealis theme — matches Kibana’s latest visual design
Collapsible sections — organize panels into groups
Export to Kibana — creates real Kibana dashboards with Lens visualizations
Import from Kibana — import existing ES|QL-based Kibana dashboards for AI-assisted editing
Custom color themes — apply custom palettes to charts, heatmaps, and metrics
Time picker — filter data by time range with automatic time field detection
Multiple dashboards — create, switch between, and manage multiple dashboards
Session isolation — parallel chat conversations work on separate dashboards via dashboardId threading
Elastic Cloud support — works with local Elasticsearch and Elastic Cloud (Cloud ID + API key)
Server instructions — workflow, tips, and capabilities exposed to every MCP client via the initialize response
Dataviz best practices — built-in guidelines for chart selection and dashboard composition
ES|QL reference — built-in language reference for correct query syntax

Architecture

For a structured walkthrough with Mermaid diagrams (system context, data flows, monorepo build order), see ARCHITECTURE.md.

Prerequisites

Node.js 22+
Elasticsearch (local or Elastic Cloud)
Kibana (for export/import)
An MCP client: Cursor (v2.6+ for MCP Apps inline preview), Claude Desktop, Claude Code, or VS Code Copilot

Quick install

在 GitHub 查看完整 README

安装

npx https://github.com/elastic/example-mcp-dashbuilder/releases/latest/download/example-mcp-dashbuilder.tgz

新手？先看 mcpskill.org 教程 →

配置

{
  "mcpServers": {
    "example-mcp-dashbuilder": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "https://github.com/elastic/example-mcp-dashbuilder/releases/latest/download/example-mcp-dashbuilder.tgz"
      ]
    }
  }
}