obsidian-playwright-mcp
Connects to Obsidian's Electron process via Chrome DevTools Protocol, enabling AI assistants to screenshot, execute JavaScript, capture console logs, and automate UI actions for debugging plugins and testing layouts.
README
obsidian-playwright-mcp
MCP server that lets AI assistants "see" and interact with the Obsidian app itself — not just read vault files.
Why This Exists
Obsidian plugins are debugged by AI today by reading source files, but that's blind. This server connects to Obsidian's Electron process via Chrome DevTools Protocol (CDP), letting Claude/any agent:
- Screenshot the app to see what's actually rendering
- Execute JavaScript in the running app context to query plugin state, trigger commands, inspect the DOM
- Capture console logs to see errors in real time
- Automate UI actions for testing plugins or validating layouts
Perfect for debugging why a plugin isn't rendering, testing UI state changes, or verifying behavior end-to-end.
Gotchas
- Obsidian must launch with
--remote-debugging-port=8315— standard startup doesn't enable CDP - Single connection at a time — only one agent can connect per Obsidian instance
- Console logs captured only after connect — startup logs are missed
- Requires Node.js v18+ and Windows (Obsidian.exe path is hardcoded)
Quick Start
-
Launch Obsidian with debug port:
Start-Process "C:\Users\patri\AppData\Local\Programs\Obsidian\Obsidian.exe" -ArgumentList "--remote-debugging-port=8315" -
Configure MCP (
.claude/mcp_config.json):{ "mcpServers": { "obsidian-automation": { "command": "node", "args": ["C:/Projects/02-Dormant-or-Ideas/obsidian-playwright-mcp/obsidian-mcp-server.js"] } } }
Tools
| Tool | Purpose |
|---|---|
connect_to_obsidian |
Establish CDP connection (run first) |
execute_in_renderer |
Run JavaScript: app.commands.executeCommandById('...'), app.vault.read(...), app.plugins.plugins[...] |
take_screenshot |
Capture current window state |
get_console_logs |
Fetch console output (filter by type: "error", "warning", etc.) |
get_page_info |
Get vault name, active file, window title |
disconnect |
Close connection |
Example: Debug a Plugin
// Connect
connect_to_obsidian()
// Returns: { success: true, title: "Home - MyVault - Obsidian v1.x" }
// Is the plugin loaded?
execute_in_renderer({ code: `!!app.plugins.plugins['my-plugin']` })
// Returns: { success: true, result: true }
// What's in the console?
get_console_logs({ filter: "error", limit: 5 })
// Returns: { logs: [{ type: "error", text: "Plugin init failed: ..." }] }
// Screenshot the error state
take_screenshot({ filename: "error-state.png" })
Tech Stack
- Node.js + Playwright (CDP connectivity)
- @modelcontextprotocol/sdk (MCP protocol)
Status: Dormant (Jan 2026). See SKILL.md for detailed examples.
Screenshots
TODO (Patrik): add screenshots here.
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。