mcp-movabletype-writer
MCP server for Movable Type. It lets MCP-compatible AI tools such as Claude Desktop work with Movable Type so that AI-generated drafts can be created and edited directly inside MT.
README
mcp-movabletype-writer
MCP server for Movable Type.
It lets MCP-compatible AI tools such as Claude Desktop work with Movable Type so that AI-generated drafts can be created and edited directly inside MT.
You can brainstorm articles with an AI assistant, store the result as a Movable Type draft, and iterate on the same draft by asking the AI to revise sections.
Features
- 🤖 AI integration: Post drafts to Movable Type straight from Claude Desktop or any MCP client.
- 💾 Session tracking: Remembers the last edited draft so multi-step rewrites stay in context.
- ✏️ Rewrite ready: “Fix this paragraph” style prompts update the current draft in place.
- 📝 Draft management: List drafts and inspect individual draft details.
For safety this server intentionally focuses on collaborative draft creation. It does not allow:
- Deleting drafts or published entries.
- Publishing drafts.
- Editing already published entries or reverting them to drafts.
Requirements
- Node.js 22.7.5 or newer (current LTS 24.x is recommended).
- Movable Type 7 r.53xx or newer with Data API enabled.
- Tested with Data API v4 and later.
Installation
If you intend to run it via
npx, you can skip this section and jump to “Using with npx”.
git clone https://github.com/burnworks/mcp-movabletype-writer.git
cd mcp-movabletype-writer
npm install
npm run build
or simply install from npm:
npm install mcp-movabletype-writer
Configuration
1. Prepare Movable Type
- In both the system dashboard and the target website/blog dashboard enable Tools → Web Services → Data API.
- Open the MT user profile that will run the API calls and note the username and Web Services Password (this is different from the regular CMS login password).
- Confirm the Data API endpoint URL (e.g.
https://example.com/your_mt_path/mt-data-api.cgi).
2. Configure Claude Desktop
Open claude_desktop_config.json.
File locations
- Windows:
%APPDATA%\Claude\claude_desktop_config.json - macOS:
~/Library/Application Support/Claude/claude_desktop_config.json
You can also open the file from Claude Desktop → Settings → Administrator.
Example configuration for a local build
Append the following block. args must contain the full path to dist/index.js.
On Windows, escape backslashes such as mcp-movabletype-writer\\dist\\index.js.
{
"mcpServers": {
"movabletype-writer": {
"command": "node",
"args": [
"/full_path_to/mcp-movabletype-writer/dist/index.js"
],
"env": {
"MT_API_URL": "https://example.com/your_mt_path/mt-data-api.cgi",
"MT_USERNAME": "your_username",
"MT_PASSWORD": "your_webservice_password",
"MT_API_VERSION": "5",
"MT_CLIENT_ID": "your_client_id",
"MT_REMEMBER": "1"
}
}
}
}
MT_API_URL: URL tomt-data-api.cgi.MT_USERNAME: Movable Type username.MT_PASSWORD: Web Services Password from the MT user profile (do not use the normal login password).MT_API_VERSION: Data API version in use, e.g.5,6,7.MT_CLIENT_ID: Any identifier composed of letters,_, or-(e.g.mcp-movabletype-writer).MT_REMEMBER:rememberflag (0/1). Leave it at1to keep sessions active until sign-out. Use0only if you need very short-lived tokens.
Changing MT_API_VERSION lets you point the same binary at MT Data API v4/v5/v6/v7 without rebuilding. MT_CLIENT_ID must be set; the server exits if it’s missing. Keeping MT_REMEMBER=1 reduces the chance of token expiry mid-session.
Using with npx
You can run the published package with npx. In that case point command to npx:
{
"mcpServers": {
"movabletype-writer": {
"command": "npx",
"args": [
"mcp-movabletype-writer"
],
"env": {
"MT_API_URL": "https://example.com/your_mt_path/mt-data-api.cgi",
"MT_USERNAME": "your_username",
"MT_PASSWORD": "your_webservice_password",
"MT_API_VERSION": "5",
"MT_CLIENT_ID": "mcp-movabletype-writer",
"MT_REMEMBER": "1"
}
}
}
}
Use mcp-movabletype-writer@1.0.0 if you prefer to pin a specific version. The env settings are identical to a local build.
Advanced settings (power users)
Fine-tuning knobs that aren’t meant for day-to-day users live in internal-config.json. Copy internal-config.example.json to the project root (or wherever you run the server from), rename it to internal-config.json, and adjust the values.
requestTimeoutMs: Timeout (milliseconds) for Movable Type HTTP requests. Defaults to30000.
If the file is missing or contains invalid JSON the server falls back to the built-in defaults, so it’s safe to experiment.
Usage
Basic flow
User: ブログID 1に「MTプラグイン開発入門」という記事の下書きを作成して
Claude: create_draftを実行...
→ 下書きを作成しました(ID: 123)
User: タイトルを「初心者向けMTプラグイン開発」に変更して
Claude: update_last_draftを実行...
→ 記事を更新しました(ID: 123)
User: 本文に「はじめに」のセクションを追加して
Claude: update_last_draftを実行...
→ 記事を更新しました(ID: 123)
Tips
- If you manage multiple blogs, ask Claude to run
list_sitesand choose the correctblog_idfrom the result. - Claude may default to HTML output; if you prefer Markdown drafts, explicitly request “save in Markdown”.
- Within a single conversation the server remembers the latest draft, so follow-up edits usually don’t require specifying
entry_id. - To edit another draft, ask for
list_recent_drafts, pick an ID from the list, and provide it toupdate_draft.
See the tool descriptions below for details.
Available tools
list_sites
Returns available blogs/sites.
Claude: list_sitesで確認...
→ ID: 1, Name: "Tech Blog"
→ ID: 2, Name: "News"
create_draft
Create a new draft.
- Required:
blog_id,title,body - Optional:
tags,categories
update_last_draft
Update the most recently created/edited draft.
- All parameters are optional; only supplied fields are changed.
update_draft
Update a draft by explicit ID.
- Required:
blog_id,entry_id - Optional:
title,body,tags,categories
get_draft
Fetch draft details.
- Required:
blog_id,entry_id
list_recent_drafts
List recent drafts.
- Required:
blog_id - Optional:
limit(default 10)
Session storage
Information about the most recent draft is stored at ~/.mcp-mt/session.json:
{
"lastEntryId": 123,
"lastBlogId": 1,
"lastUpdated": "2025-11-07T10:00:00.000Z"
}
This lets update_last_draft run without specifying entry_id.
For developers
Using environment variables with npm run dev
If you want to iterate with npm run dev (tsx) instead of npm run build, copy the example env file:
cp .env.example .env
MT_API_URL=https://example.com/your_mt_path/mt-data-api.cgi
MT_USERNAME=your_username
MT_PASSWORD=your_webservice_password
MT_API_VERSION=5
MT_CLIENT_ID=your_client_id
MT_REMEMBER=1
Then run:
npm install
npm run dev
For debugging, tools like @modelcontextprotocol/inspector make it easy to connect and exercise the MCP server while you develop.
Troubleshooting
Authentication errors
- Ensure
MT_USERNAME,MT_PASSWORD(Web Services Password), andMT_CLIENT_IDare correct. - Verify the Data API is enabled and that
mt-data-api.cgiis accessible.
Cannot find drafts
- Use
list_sitesto confirm the correctblog_id. - Run
list_recent_draftsto see available drafts and their IDs.
Session keeps resetting
- Check that
~/.mcp-mt/session.jsonstill exists. - Restarting Claude Desktop starts a new MCP session (and thus a blank
session.json).
License
MIT
References
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。