io.github.gaopengbin/cesium-mcp-dev

Demo

https://github.com/user-attachments/assets/8a40565a-fcdd-47bf-ae67-bc870611c908

Packages & Entry Points

Module	Role	Status	Links
cesium-mcp-bridge	Protocol-agnostic command dispatcher (60+ tools)	Mainline, actively iterated	· source
examples/browser-agent	Browser-only AI agent (recommended starting point, zero backend)	Recommended	example · live demo
cesium-mcp-runtime	MCP server (stdio + HTTP)	Stable, slow updates	· source
cesium-mcp-dev	CesiumJS API knowledge base for coding assistants	Maintained	· source

Which one? Personal project or quick try → browser-agent. Existing web app embedding an AI assistant → bridge + your own function calling. Calling from Claude Desktop / Cursor / Dify → MCP runtime.

Architecture

flowchart LR
  subgraph clients ["AI Drivers (pick one)"]
    BA["Browser Agent\n(in the same page)"]
    FC["Your web app\nfunction calling"]
    MCP["Claude / Cursor / Dify\nvia MCP runtime"]
  end

  subgraph core ["cesium-mcp-bridge (browser)"]
    B["60+ tools\nprotocol-agnostic dispatcher"]
    C["CesiumJS Viewer"]
  end

  BA -- "in-page call" --> B
  FC -- "in-page call" --> B
  MCP -- "WebSocket / JSON-RPC" --> B
  B --> C

  style clients fill:#1e293b,stroke:#528bff,color:#e2e8f0
  style core fill:#1e293b,stroke:#12B76A,color:#e2e8f0

The bridge is the only required piece. Pick whichever driver matches your scenario — they all hit the same 60+ tools.

Quick Start

Path 0 — Try in 30 seconds (browser agent, recommended)

Open the live demo, paste an OpenAI-compatible API key, and ask:

"Fly to the Eiffel Tower and drop a red marker"

Fork the examples/browser-agent folder to deploy your own.

Path 1 — Embed in your own web app (function calling)

npm install cesium-mcp-bridge

import { CesiumBridge } from 'cesium-mcp-bridge';

const bridge = new CesiumBridge(viewer);
// Then: send the bridge's tool schema to any LLM that supports function/tool calling,
// route the model's tool calls to bridge.execute(name, params).

See examples/browser-agent/index.html for a complete loop with OpenAI-compatible APIs.

Path 2 — Use from Claude Desktop / Cursor / Dify (MCP)

Install bridge as in Path 1, then start the MCP runtime:

# stdio mode (Claude Desktop, VS Code, Cursor)
npx cesium-mcp-runtime

# HTTP mode (Dify, remote/cloud MCP clients)
npx cesium-mcp-runtime --transport http --port 3000

MCP client config:

{
  "mcpServers": {
    "cesium": {
      "command": "npx",
      "args": ["-y", "cesium-mcp-runtime"]
    }
  }
}

58 Available Tools

Tools are organized into 12 toolsets. Default mode enables 4 core toolsets (~31 tools). Set CESIUM_TOOLSETS=all for everything, or let the AI discover and activate toolsets dynamically at runtime.

i18n: Tool descriptions default to English. Set CESIUM_LOCALE=zh-CN for Chinese.

Relationship with CesiumGS official MCP servers: The camera, entity-ext, and animation toolsets natively fuse capabilities from CesiumGS/cesium-mcp-server (Camera Server, Entity Server, Animation Server) into this project's unified bridge architecture. This means you get all official functionality plus additional tools — in a single MCP server, without running multiple processes.

Examples

See examples/minimal/ for a complete working demo.

Development

git clone https://github.com/gaopengbin/cesium-mcp.git
cd cesium-mcp
npm install
npm run build

Version Policy

Version format: {CesiumMajor}.{CesiumMinor}.{MCPPatch}

When CesiumJS releases a new minor version (e.g. 1.140), we will bump accordingly: 1.140.0.

Related Projects

• mapbox-mcp — AI control for Mapbox GL JS
• openlayers-mcp — AI control for OpenLayers

Star History

License

MIT