Woow VK MCP Server

Woow VK MCP Server

Three-piece MCP service for Vibe Kanban that exposes 33 tools via authenticated SSE/StreamableHttp, enabling AI agents to manage workspaces, repositories, issues, and Kanban boards.

Category
访问服务器

README

<p align="center"> <img src="docs/screenshots/mcp-02-dashboard.png" alt="VK MCP Admin Dashboard" width="720"/> </p>

<h1 align="center">Woow VK MCP Server</h1>

<p align="center"> <strong>WoowTech Three-Piece MCP Service for Vibe Kanban</strong><br/> Proxy + Admin GUI + Supergateway Bridge — 33 VK Tools via Authenticated SSE/StreamableHttp </p>

<p align="center"> <a href="#architecture">Architecture</a> • <a href="#tools">Tools</a> • <a href="#screenshots">Screenshots</a> • <a href="#deployment">Deployment</a> • <a href="#connection">Connection</a> • <a href="README_zh-TW.md">中文文件</a> </p>

<p align="center"> <img src="https://img.shields.io/badge/MCP-33_Tools-blue" alt="33 Tools"/> <img src="https://img.shields.io/badge/Transport-StreamableHttp-green" alt="StreamableHttp"/> <img src="https://img.shields.io/badge/Vibe_Kanban-v0.1.44-purple" alt="VK"/> <img src="https://img.shields.io/badge/Pattern-Three--Piece_MCP-orange" alt="Pattern"/> <img src="https://img.shields.io/badge/License-MIT-green" alt="License"/> </p>


Overview

A production-ready MCP (Model Context Protocol) service that exposes 33 Vibe Kanban tools to external AI agents (Claude, n8n, custom) via authenticated HTTPS endpoints. Follows the WoowTech standard three-piece MCP pattern proven across K3s, Odoo, Hermes, and n8n deployments.

How It Works

Claude App / AI Agent
  → HTTPS (Cloudflare Tunnel)
    → Proxy (token auth + reverse proxy)
      → Supergateway (stdio → StreamableHttp bridge)
        → vibe-kanban-mcp binary (33 tools, Rust)
          → VK Host API (localhost:3000, same Pod)
            → Self-hosted VK Remote (issues/kanban)
            → Self-hosted VK Relay (host connections)

Architecture

Three-Piece Pattern

graph TB
    subgraph "External"
        Agent[AI Agent<br/>Claude / n8n / Custom]
    end

    subgraph "K8s Pod: vk-host"
        subgraph "Container: mcp-service"
            Proxy["件1: Proxy<br/>/private_{token}/*<br/>Token Auth + SSE Forward"]
            Admin["件2: Admin GUI<br/>FastAPI + React SPA<br/>:8080"]
            SG["件3: Supergateway<br/>stdio → StreamableHttp<br/>:8000"]
            MCP["vibe-kanban-mcp<br/>--mode global<br/>33 tools (Rust binary)"]
        end

        subgraph "Container: host"
            Host["VK Host Server<br/>:3000<br/>Workspace + Agent Executor"]
        end
    end

    subgraph "Other Pods"
        Remote["VK Remote<br/>Web UI + API<br/>:8081"]
        Relay["VK Relay<br/>WebSocket<br/>:8082"]
        DB[(PostgreSQL 16)]
        Electric[ElectricSQL]
    end

    Agent -->|HTTPS| Proxy
    Proxy --> SG
    SG -->|stdio| MCP
    MCP -->|localhost:3000| Host
    Host --> Remote
    Host --> Relay
    Remote --> DB
    Electric --> DB

Sidecar Deployment

The MCP service runs as a sidecar container in the same Pod as the VK Host. They share a network namespace, enabling localhost:3000 direct access without authentication (Host has no auth middleware — security boundary is the Pod network).

┌──────────────── K8s Pod: vk-host ────────────────┐
│                                                    │
│  Container: mcp-service        Container: host     │
│  ┌─────────────────────┐      ┌──────────────────┐│
│  │ Proxy    :8080       │      │ VK Host   :3000  ││
│  │ Gateway  :8000       │◄────►│ Claude Code      ││
│  │ vibe-kanban-mcp      │      │ OpenCode         ││
│  └─────────────────────┘      └──────────────────┘│
│       localhost network (shared)                    │
└────────────────────────────────────────────────────┘

Tools

33 MCP Tools in 4 Categories

Category Count Tools
A - Execution 10 start_workspace, create_session, run_session_prompt, get_execution, list_sessions, update_session, list_workspaces, update_workspace, delete_workspace, link_workspace_issue
B - Repo 5 list_repos, get_repo, update_setup_script, update_cleanup_script, update_dev_server_script
C - Issue/Kanban 17 list_organizations, list_org_members, list_projects, create_issue, list_issues, get_issue, update_issue, delete_issue, list_issue_priorities, assign_issue, unassign_issue, list_issue_assignees, add_issue_tag, remove_issue_tag, list_issue_tags, list_tags, create_issue_relationship, delete_issue_relationship
D - Context 1 get_context

Dangerous Tools (destructive operations)

Tool Risk
delete_workspace Permanently deletes workspace + worktree
delete_issue Permanently deletes an issue
start_workspace Creates worktrees on host filesystem

Note: This proxy-upstream architecture exposes all tools. Selective blocking requires the future own-server migration. See deny-list.yaml.


Screenshots

Login Page

<p align="center"> <img src="docs/screenshots/mcp-01-login.png" alt="Login" width="720"/> </p>

Dashboard

System health overview showing MCP Server, Proxy, and Tunnel status. <p align="center"> <img src="docs/screenshots/mcp-02-dashboard.png" alt="Dashboard" width="720"/> </p>

Tools (33 VK MCP Tools)

Complete tool registry organized by category with danger flags. <p align="center"> <img src="docs/screenshots/mcp-03-tools.png" alt="Tools" width="720"/> </p>

Token Management

Generate, rotate, and manage MCP authentication tokens. <p align="center"> <img src="docs/screenshots/mcp-04-tokens.png" alt="Tokens" width="720"/> </p>

Log Viewer

Real-time supergateway and proxy logs. <p align="center"> <img src="docs/screenshots/mcp-05-logs.png" alt="Logs" width="720"/> </p>

Settings

MCP server configuration and connection settings. <p align="center"> <img src="docs/screenshots/mcp-06-settings.png" alt="Settings" width="720"/> </p>


Deployment

Prerequisites

  • K3s / Kubernetes cluster with existing VK stack (PostgreSQL, ElectricSQL, Remote, Relay, Host)
  • podman or docker for building images
  • Cloudflare account (for tunnel)

Build

# Build the MCP service image
podman build -t vk-mcp:v1.0 .

# Import to K3s
podman save localhost/vk-mcp:v1.0 -o /tmp/vk-mcp.tar
# (use privileged Job to import into containerd)

K8s Deployment (Sidecar)

Add the mcp-service container to the existing vk-host Pod:

containers:
  - name: host          # existing
    image: localhost/vk-host:v0.1.43
    ports: [{containerPort: 3000}]
  - name: mcp-service   # NEW sidecar
    image: localhost/vk-mcp:v1.0
    ports:
      - {containerPort: 8080, name: mcp-admin}
      - {containerPort: 8000, name: mcp-sse}
    env:
      - {name: VIBE_BACKEND_URL, value: "http://localhost:3000"}
      - {name: MCP_AUTH_TOKEN, value: "<your-token>"}

Cloudflare Tunnel

Add route: your-mcp-domain.example.com → vk-host-svc:8080


Connection

Claude App / Claude Desktop

https://<your-mcp-domain>/private_<token>/mcp

Claude Code / Settings JSON

{
  "mcpServers": {
    "vibe-kanban": {
      "type": "url",
      "url": "https://<your-mcp-domain>/private_<token>/mcp"
    }
  }
}

Token Management

  1. Login to Admin GUI: https://<your-mcp-domain>/ (password: admin)
  2. Navigate to Tokens page
  3. Click Rotate Token to generate a new token
  4. Use the token in your MCP connection URL

Project Structure

woow_vk_mcp_server/
├── mcp_admin_core/          # 件1: Shared MCP proxy + auth (zero-change copy)
│   ├── proxy.py             # SSE/StreamableHttp reverse proxy with token auth
│   ├── process.py           # Subprocess manager for supergateway
│   ├── app.py               # FastAPI factory with SPA fallback
│   ├── auth/                # JWT middleware
│   └── config/              # JSON config store
├── vk_mcp_admin/            # 件2: VK-specific admin backend
│   ├── main.py              # create_app("VK MCP Admin")
│   ├── tool_registry.py     # 33 tools in 4 categories
│   └── routers/             # health, tools, tokens, logs, config
├── frontend/                # 件2: React SPA (Vite + TailwindCSS)
│   └── src/pages/           # Dashboard, Tools, Tokens, Logs, Settings, Login
├── Dockerfile               # Multi-stage: Node 22 + Python 3.12 + binary
├── entrypoint.sh            # Starts supergateway + uvicorn
├── deny-list.yaml           # Warning: all 33 tools exposed
├── .env.example             # Configuration template
└── pyproject.toml            # Python dependencies

Test Results (33/33 = 100%)

All 33 MCP tools verified with full create → update → delete closed-loop testing:

Category Tools Status
C - Issue/Kanban 18 tools 18/18 PASS
B - Repo 5 tools 5/5 PASS
A - Execution 10 tools 10/10 PASS
Total 33 tools 33/33 (100%)

Live Instance

Service URL
MCP Admin GUI https://woowtechkxs-vibekanban-mcp.woowtech.io
MCP Endpoint https://woowtechkxs-vibekanban-mcp.woowtech.io/private_{token}/mcp
VK Remote (Web UI) https://woowtechkxs-vibekanban.woowtech.io
VK Host https://woowtechkxs-vibekanban-host.woowtech.io

Related


License

MIT License - see LICENSE for details.


<p align="center"> <sub>Built by <a href="https://github.com/WOOWTECH">WOOWTECH</a> | Powered by <a href="https://github.com/BloopAI/vibe-kanban">Vibe Kanban</a> + <a href="https://modelcontextprotocol.io">MCP</a></sub> </p>

推荐服务器

Baidu Map

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。

官方
精选
JavaScript
Playwright MCP Server

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。

官方
精选
TypeScript
Magic Component Platform (MCP)

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。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

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。

官方
精选
本地
TypeScript
VeyraX

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。

官方
精选
本地
graphlit-mcp-server

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。

官方
精选
TypeScript
Kagi MCP Server

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。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

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。

官方
精选