mcp-divoom-lan
Enables AI clients to control Divoom LAN devices, including watchface management, brightness adjustment, and screen control through natural language.
README
mcp-divoom-lan
mcp-divoom-lan is an open-source MCP server that wraps Divoom watchface LAN APIs as standard tools for AI clients.
It works together with the v2 HTML visual editor for modifying watchfaces, switching faces, adjusting brightness, and creating new local watchfaces.
v2 visual editor (public):
- GitHub:
https://github.com/DivoomDevelop/divoom-watchface-visual-editor_v2 - Live site:
https://divoomdevelop.github.io/divoom-watchface-visual-editor_v2/
Your local clone path (e.g. D:\divoom-watchface-visual-editor) is machine-specific; use the v2 GitHub / GitHub Pages URLs above in docs and MCP metadata.
Goals
- Expose key capabilities from
Divoom_Watchface_Remote_Customization_Guide_EN.mdas MCP tools - Let MCP-enabled clients (Cursor, Claude Desktop, local LLMs, etc.) drive watchface actions via natural language
- Preserve safety boundaries (read before write, explicit warnings for risky operations, multipart rules)
Default safety policy (important)
- Read before write: call
watchface_get_local, thenwatchface_patch_local, then read back to verify. - If
GetLocalClockInforeturns an emptyItemList: stop writes; switch to an editable watchface first. - Do not call
watchface_create_local_clockunless the user clearly asks to create a new one (no implicit creation).
Implemented tools
watchface_get_local→Device/GetLocalClockInfowatchface_patch_local→Device/PatchLocalClockInfo(default/divoom_api); optionaldialAssetsPathswitches to multipartPOST /patch_local_clock(same dial/tar.gz rules aswatchface_create_local_clock)watchface_get_fonts_local→Device/GetLocalFontListwatchface_get_store_market_list→Device/GetStoreClockMarketListwatchface_set_clock_select→Channel/SetClockSelectIdwatchface_get_brightness→Sys/GetBrightnesswatchface_set_brightness→Channel/SetBrightnesswatchface_onoff_screen→Channel/OnOffScreen(1=on, 0=off)watchface_replace_dial_bg_file→POST /replace_clock_dial_bgwatchface_upload_file→POST /uploadwatchface_create_local_clock→POST /create_local_clock(multipart: single dial image ortar.gz; JSONDialAssets/UseDialAssetBundleselects mode, default auto-detect gzip)watchface_reset_local_then_cloud→Device/ResetLocalClockFromServerwatchface_raw_command→ genericPOST /divoom_apiwatchface_protocol_quick_reference→ key protocol constraints for the model
Resources (context for the model)
The server exposes two MCP resources:
divoom://guide/quick-referencedivoom://skill/watchface-customization
MCP Bundle (.mcpb)
For MCPB-compatible hosts (e.g. Claude desktop connectors, Smithery stdio releases), build a local bundle:
- Install the packer:
npm install -g @anthropic-ai/mcpb - From this package root:
npm run mcpb:pack - Output:
mcp-divoom-lan.mcpb(gitignored). The staging directorymcpb/staging/is also gitignored.
The bundle includes dist/, resources/, production node_modules, and a manifest.json with user fields for device IP, port, and timeout.
Quick start
cd tools/mcp-divoom-lan # or your clone root for this package
npm install
npm run build
npm start
Development (watch rebuild):
npm run dev
Pre-release check (typecheck, build, pack dry-run):
npm run release:check
Documentation
docs/README.md— documentation indexdocs/quick-start.md— minimal setupdocs/tool-examples.md— tool usage examples (includes §5b analog pointer layout)docs/disp-usage.md— choosingdispids (pointer layout131/132/233; net-gallery uniqueness13/125–130/173–175)docs/html-visual-editor.md— using the visual editor with MCPdocs/safety-and-troubleshooting.md— safety and FAQsdocs/reference/— condensed protocol rules (EN/ZH)docs/examples/— sample requests/responses and catalog
Environment variables
DIVOOM_DEVICE_HOST— device LAN IP (e.g.192.168.1.120)DIVOOM_DEVICE_PORT— HTTP port, default9000DIVOOM_TIMEOUT_MS— request timeout ms, default45000
If DIVOOM_DEVICE_HOST is unset, each tool call must pass target.host.
Example client config (stdio)
Cursor / Claude Desktop
{
"mcpServers": {
"divoom-lan": {
"command": "node",
"args": [
"/ABSOLUTE/PATH/to/tools/mcp-divoom-lan/dist/index.js"
],
"env": {
"DIVOOM_DEVICE_HOST": "192.168.1.120",
"DIVOOM_DEVICE_PORT": "9000",
"DIVOOM_TIMEOUT_MS": "45000"
}
}
}
}
You can also copy client-config.example.json in this directory as a starting point.
Publishing checklist (for maintainers)
- Use a dedicated repo (e.g.
mcp-divoom-lan) with this package at the repo root. - Verify metadata:
LICENSE,SECURITY.md,CONTRIBUTING.md,CHANGELOG.md,RELEASE.mdas applicable. - Run
npm run release:check. - Tag a GitHub release (e.g.
v0.1.2) with screenshots and sample requests if helpful. - Submit listings where appropriate (MCP Registry, Smithery, Glama, MCP.so, community indexes). For Glama, follow
GLAMA_SUBMISSION_READY.md(includesDockerfileandglama.json). For MCP.so, followMCP_SO_SUBMISSION_READY.md. For 火山引擎 MCP 清单,见VOLCENGINE_SUBMISSION_READY.md(PR: https://github.com/volcengine/mcp-server/pull/398)。For 阿里云百炼自定义 MCP(控制台 npx 部署),见BAILIAN_MCP_SUBMISSION_READY.md。For 扣子 Coze 插件发布/商店(HTTP 插件,与 MCP 不同),见COZE_SUBMISSION_READY.md。 - Minimal demo flow:
watchface_get_local→watchface_patch_local(font size/color) →watchface_replace_dial_bg_file(background).
Files often used at release
Included in this repo (when present): LICENSE, CHANGELOG.md, CONTRIBUTING.md, SECURITY.md, RELEASE.md, optional checklist and directory templates, and .github/workflows/ci.yml.
Should the HTML visual editor ship inside this npm package?
Recommendation: no for the core MCP package — keep MCP lean. Offer the editor as a separate optional project.
- Core:
https://github.com/DivoomDevelop/mcp-divoom-lan - Visual editor v2:
https://github.com/DivoomDevelop/divoom-watchface-visual-editor_v2 - Hosted v2:
https://divoomdevelop.github.io/divoom-watchface-visual-editor_v2/
Benefits:
- Small MCP install suitable for all AI clients
- Non-developers can use the visual UI to understand
ItemList, then let the AI apply patches - Clear split between WYSIWYG editing and automated MCP writes
Alignment with upstream docs
This repo ships standalone docs under docs/, docs/reference/, and docs/examples/. If you maintain full guides elsewhere, keep this tree synced or treat it as the distribution subset.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。