superthread-mcp-extended
An enhanced MCP server for Superthread with comprehensive CRUD operations and optimized workflows, acting as a drop-in replacement with more functionality.
README
Superthread MCP Extended
A tiny MCP server by Sorbet Studio
An enhanced Model Context Protocol (MCP) server that extends the official Superthread MCP with comprehensive CRUD operations and optimized workflows. Drop-in replacement with significantly more functionality.
Chose cloudflare workers over vercel since this app is pretty much serverless functions, with no complex nodejs features being uses, just simple api request tranforms. Besides, I love cloudflare's free tier.
Please help create an issue/PR if you wanna add more tools or improve the existing ones.
README is a work in progress. Ideas and thoughts are scattered rn.
Key Difference from the official MCP
- 8 essential tools instead of basic ask/search functionality
- Complete CRUD operations for cards, boards, and spaces
- Optimized responses for reduced token usage and faster tool chaining .. compared to raw api.
Quick Start
Use our hosted version directly without setup, with zero logging or data retention:
https://superthread.sorbet.studio/mcp/app
Note: The Cloudflare worker for this, literally just builds this main branch and deploys it. If you find any security or privacy loop holes, please open an issue, and i'll fix it.
The idea is to just create a perfect drop in replacement to the original, no other code should ideally run apart from transforming/filtering the api requests from superthread. So you wont have to create your own instance, but if you want to, you should! (Deploy to Cloudflare button coming soon.)
Here's the MCP setup guide from api.superthread.com/mcp .. just the domain swapped in.
Claude Desktop
Add to your claude_desktop_config.json:
{
"mcpServers": {
"superthread-extended": {
"command": "npx",
"args": [
"-y",
"mcp-remote@latest",
"https://superthread.sorbet.studio/mcp/app",
"--header",
"Authorization:${ST_PAT}"
],
"env": {
"ST_PAT": "Bearer stp-XXXXXXXXXXXX.XXXXXXXXXXXXXXXXXXXXXXXX"
}
}
}
}
Claude Code
claude mcp add \
--transport http \
--scope local \
"superthread-extended" \
"https://superthread.sorbet.studio/mcp/app" \
--env ST_PAT="stp-XXXXXXXXXXXX.XXXXXXXXXXXXXXXXXXXXXXXX" \
--header "Authorization: Bearer ${ST_PAT}"
Cursor
Add to ~/.cursor/mcp.json:
{
"mcpServers": {
"superthread-extended": {
"type": "http",
"url": "https://superthread.sorbet.studio/mcp/app",
"headers": {
"Authorization": "Bearer stp-XXXXXXXXXXXX.XXXXXXXXXXXXXXXXXXXXXXXX"
}
}
}
}
Available Tools
User Management
get_my_account- Get team ID and user informationget_team_members- List team members
Workspace Navigation
get_spaces- List all spaces/projects with project IDsget_space- Get detailed space information
Board Management
get_board- Search for boards by titleget_board_details- Get board with lists for task creationget_boards- List boards with filtering optionscreate_board- Create new boards (admin/owner only)update_board- Update board properties
Task Management
create_card- Create tasks from any inputget_card- Get detailed task informationupdate_card- Update task propertiesget_cards_assigned_to_user- Get user's assigned tasksadd_related_card- Link related tasksarchive_card- Archive completed tasks
Core Workflow
A typical 4-step process for say, extracting tasks from a screenshot using claude:
get_my_account()→ Extract team_idget_spaces(team_id)→ Get project_id and available boardsget_board_details(team_id, board_id)→ Get list_id for task placementcreate_card(...)→ Create tasks with proper hierarchy
Again, as MCP servers go, you could let Claude (etc) figure the tool call chain.
Architecture
Response Optimization
- Intelligent field filtering removes 60-80% of unnecessary data
- Preserves essential IDs and relationships for tool chaining
- Maintains full API compatibility with reduced token usage
Security Boundaries
- Read operations: Full access to user's workspace data
- Create operations: Cards and boards only (with permissions)
- Excluded: Destructive operations, member management, space creation
Project Structure
src/
├── lib/
│ ├── api-client.ts # Centralized Superthread API client
│ ├── response-filters.ts # Response optimization
│ └── search.ts # Search utilities
├── tools/
│ ├── user.ts # User and team operations
│ ├── spaces.ts # Workspace navigation
│ ├── boards.ts # Board management
│ └── cards.ts # Task operations
├── types/ # TypeScript definitions
├── index.ts # Cloudflare Workers entry
└── server.ts # MCP server configuration
Self-Hosting
Development
git clone <repository>
cd superthread-mcp-extended
bun install
bun run dev
Deployment
# Deploy to Cloudflare Workers
bun run deploy
Local Usage
http://localhost:8787/mcp/app
Requirements
- Superthread Personal Access Token (PAT)
- Node.js 22+ or Bun runtime (for development)
- Network access to api.superthread.com
Authentication
All requests require a valid Superthread Personal Access Token:
Authorization: Bearer stp-XXXXXXXXXXXX.XXXXXXXXXXXXXXXXXXXXXXXX
Tokens can be generated in your Superthread account settings.
Note: This is an unofficial extension and is not affiliated with Superthread. For the official basic MCP server, visit api.superthread.com/mcp.
Credit:
- Superthread for the original MCP server and fantastic API docs (ofc the platform as well haha)
- Cloudflare for the Workers platform
- Unofficial-Superthread-MCP for the initial inspiration and some code snippets, and the idea of using ZOD for request validation. Really cool project, please check them out too as a direct local alternative to this one.
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。