pen-editor
Provides MCP tools to read and edit live Pen Editor documents, including structure, layout, variables, screenshots, and batch design DSL, plus static guidelines and style guides.
README
pen-editor-plugin
Pen Editor packaged as an Agent Plugins 1.0.0 plugin, installable in any client that supports the specification (VS Code, Cursor, and others).
Installing it gives the agent:
- the
pen-editorMCP server — read and edit a live Pen Editor document: structure, computed layout, design variables, screenshots, and thebatch_designDSL for creating and modifying designs; - three skills the client picks up on its own, no prompt wiring needed:
pen-editor-design(build or edit a design),pen-editor-dev-mode(implement an existing design as code), andpen-editor-setup(diagnose a broken connection).
Setup
Local development (zero-config)
- Install the plugin.
- Run the backend:
With nocd pen-editor-backend npm run devMCP_AUTH_TOKENset, the backend generates one itself, serves MCP loopback-only, and writes it to~/.pen-editor/mcp.json. - Run the editor:
It picks up the same token automatically.cd pen-editor npm run dev - Done — the plugin discovers the backend on its own. Most tools need the
editor tab open in a browser;
get_guidelines,get_style_guide_tagsandget_style_guidework without one.
Remote backend
For a backend that isn't running on localhost, or one with an explicit
MCP_AUTH_TOKEN, point the plugin at it:
cd pen-editor-backend
MCP_AUTH_TOKEN=<secret-at-least-16-chars> npm run dev
cd pen-editor
VITE_MCP_WS_TOKEN=<same-secret> npm run dev
Note: VITE_MCP_WS_TOKEN is inlined into the public JS bundle at build
time. Use it only in a local/dev build — never on a publicly deployed
frontend, or any visitor gets the secret.
Then tell the plugin the URL and token, either by asking the agent to run
its configure_pen_editor_connection tool (paste the URL and token into
chat), or by writing config.json into the plugin's data directory (the
client decides the exact path):
{
"url": "https://my-backend.example.com/api/mcp",
"token": "<secret-at-least-16-chars>"
}
Environment variables PEN_EDITOR_MCP_URL / PEN_EDITOR_MCP_TOKEN are also
supported as a power-user channel — but it's non-portable (not every client
preserves ambient env vars) and lower precedence than
<PLUGIN_DATA>/config.json, so once configure_pen_editor_connection has
written that file, these env vars no longer have any effect.
License
MIT — see LICENSE.
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。