openart-mcp
Automates OpenArt.ai character and video workflows using Playwright, exposing tools to list, create characters, generate videos, and check render status.
README
openart-mcp
MCP server + Claude skill for OpenArt.ai. OpenArt has no public API, so this drives their web UI as a logged-in user via Playwright.
What's in this repo
- MCP server (
src/) — exposes 5 tools that automate OpenArt character + video workflows - Claude skill (
skills/openart/SKILL.md) — tells Claude when and how to call the MCP tools
Tools
| Tool | Purpose |
|---|---|
openart_list_characters |
List every character in your OpenArt account |
openart_get_character |
Get details for one character by ID |
openart_create_character |
Create a new character from a local image |
openart_generate_video |
Generate a video of a character speaking a script |
openart_get_video_status |
Check render status + URL of a generated video |
Install
1. Install the MCP server
git clone https://github.com/jbertus/openart-mcp.git
cd openart-mcp
npm install
npx playwright install chromium
npm run build
npm run login # opens browser, log in to OpenArt, press Enter when done
The login step saves a Playwright storageState JSON to ~/.openart-mcp/auth.json (or %USERPROFILE%\.openart-mcp\auth.json on Windows). The server reuses that session for every subsequent tool call.
2. Register the MCP with your client
Claude Code / Claude Desktop — add to your MCP config (~/.claude/mcp.json or equivalent):
{
"mcpServers": {
"openart": {
"command": "node",
"args": ["<absolute-path-to-cloned-repo>/dist/index.js"]
}
}
}
Other MCP clients — point to node <repo>/dist/index.js with stdio transport.
3. Install the companion skill (optional but recommended)
npx skills add https://github.com/jbertus/openart-mcp --skill openart
The skill teaches Claude when to call which tool and how to format arguments. Without it, Claude can still use the MCP but you'll have to spell out intent more explicitly.
Configuration
Environment variables (all optional):
| Var | Default | Purpose |
|---|---|---|
OPENART_STORAGE_STATE |
~/.openart-mcp/auth.json |
Path to Playwright session JSON |
OPENART_HEADLESS |
true |
Set to false to see the browser during automation |
Selector setup (one-time)
The tools in src/tools.ts ship with placeholder DOM selectors marked TODO. OpenArt's UI changes over time, so the selectors need to be captured against the live site:
npx playwright codegen https://openart.ai/suite/characters-and-worlds
Click through the UI for each tool's flow. Playwright generates real selectors. Paste them into the TODO spots in src/tools.ts, rebuild (npm run build), and the tools work.
If the selectors break later (OpenArt redesigns), repeat this step.
Risks
- Fragile to UI changes. Every OpenArt redesign may break selectors. Re-run codegen and re-patch.
- Slower than a real API. Each tool call launches a browser session and waits on page loads.
- Session can expire. Re-run
npm run loginwhen prompted. - OpenArt may block automation. Anti-bot detection could rate-limit or ban your account. Use sparingly. No recourse if banned.
Testing
npm run inspect # launches MCP Inspector against the server
Contributing
PRs welcome. If OpenArt's UI changes and the selectors break, open an issue with the new selectors and a brief description of which tool flow needs updating.
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。