Anytype MCP Server
Enables AI assistants to interact with Anytype data through the local API, allowing search, creation, modification, and export of objects, spaces, and lists.
README
OFFICIAL ANYTYPE MCP WAS RELEASED HERE
https://github.com/anyproto/anytype-mcp
Anytype MCP Server
An MCP (Model Context Protocol) server that provides access to the Anytype API, allowing AI assistants and other MCP clients to interact with your Anytype data.
Based on the Anytype API definition v0.46+ (2025-03-17). Ensure your Anytype Desktop version is compatible.
Features
- Get list of spaces (
get_spaces) - Search/Get objects within a space (
get_objects,search_space) or globally (global_search) - Get detailed object content (
get_object_content, supports retrieving full text) - Create and delete spaces (
create_space) and objects (create_object,delete_object) - Export objects as markdown (
export_object) - Manage list views and objects within lists (
get_list_views,get_list_view_objects,add_objects_to_list,remove_object_from_list) - Get space members (
get_space_members) - Get types and templates (
get_types,get_type_details,get_templates,get_template_details)
Prerequisites
- Node.js 18 or higher
- Anytype desktop application running locally
- An Anytype account
Installation
Manual Installation
-
Clone this repository:
git clone https://github.com/Qwinty/anytype-mcp.git cd anytype-mcp -
Install dependencies:
npm install -
Build the project (compiles TypeScript to JavaScript in
build/):npm run build -
Obtain an App Key: Before configuring the server, you need an App Key from Anytype. See the "Getting an App Key" section below.
-
Add the MCP server to your MCP configuration file
Getting an App Key
Before using the MCP server, you need to obtain an app key from the Anytype desktop application:
-
Make sure Anytype desktop is running
-
Run the helper script:
npm run get-key -
Follow the instructions to authorize the application
-
Note the app key for configuration
Configuration
Add the Anytype MCP server to your MCP configuration file:
- For Claude: Edit
claude_desktop_config.json - For other MCP clients: Edit their respective configuration files
Example configuration:
{
"mcpServers": {
"anytype": {
"command": "node",
"args": ["{path/to/anytype-mcp}/build/index.js"],
"env": {
"ANYTYPE_APP_KEY": "YOUR_APP_KEY_HERE"
},
"disabled": false
}
}
}
Replace path/to/anytype-mcp with the actual path to your installation and YOUR_APP_KEY_HERE with the app key you obtained.
Usage
Starting the server
The MCP server is usually started automatically by the MCP client. However, you can also start it manually for testing:
npm start
Available Tools
See Tools.md for a detailed list of available tools and their usage examples.
System Prompt
Additionally, a sample system prompt for AI assistants using this server is available in docs/system-prompt.md.
Token Efficiency and Data Filtering
To optimize for token usage with AI assistants, this MCP server implements response filtering by default for tools that return object data (get_objects, global_search, search_space).
- Default Behavior: The server returns a simplified version of object data, including essential metadata like ID, name, type, icon, layout, space ID, root ID, snippet (if available), block count, tags, creation/modification dates, and creator info. Full block content and detailed relations are omitted.
include_text: true: Several tools (get_objects,get_object_content,global_search,search_space) support an optionalinclude_textparameter. When set totrue, the server will extract and include the full, formatted text content from the object's blocks in afull_textfield. Use this when you need the complete text, but be aware it significantly increases response size and token count.full_response: true: Theget_objects,global_search, andsearch_spacetools also support afull_responseparameter. Setting this totruebypasses all filtering and returns the raw, complete JSON response directly from the Anytype API. This provides the most detail but uses the most tokens.
Choose the appropriate parameters based on whether you need just metadata, full text content, or the complete raw API response.
Troubleshooting
Anytype API Not Responding
Make sure the Anytype desktop application is running on your computer. The MCP server connects to the local Anytype API at http://localhost:31009/v1.
Authentication Issues
If you encounter authentication errors:
- Run
npm run get-keyto obtain a new app key - Update your MCP configuration with the new key
- Restart your MCP client
Local API Port
The server connects to the Anytype API at http://localhost:31009/v1 by default. If your Anytype installation uses a different port, you currently need to modify the apiBaseUrl variable in src/index.ts and rebuild (npm run build). Making this configurable via an environment variable is a potential future improvement.
License
MIT
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。