Perdoo MCP Server
An MCP server that integrates with the Perdoo OKR platform, providing tools to manage objectives, key results, initiatives, KPIs, and strategic pillars.
README
Perdoo MCP Server
An MCP (Model Context Protocol) server that integrates with the Perdoo OKR platform via its GraphQL API. Provides 19 tools for managing objectives, key results, initiatives, KPIs, and strategic pillars.
Features
- Full CRUD for objectives, key results, initiatives, KPIs, and strategic pillars
- Helper tools for looking up timeframes, users, and groups
- Cursor-based pagination on all list operations
- Resilience stack: rate limiting, circuit breaker, retry with exponential backoff, request queuing
- Session-based HTTP transport (Streamable HTTP)
- Instructions resource at
perdoo://instructionsfor LLM guidance
Tools
| Category | Tools |
|---|---|
| Objectives | list_objectives, get_objective, create_objective, update_objective |
| Key Results | list_key_results, get_key_result, create_key_result, update_key_result |
| Initiatives | list_initiatives, get_initiative, create_initiative, update_initiative |
| KPIs | list_kpis, get_kpi, create_kpi, update_kpi |
| Strategic Pillars | list_strategic_pillars, get_strategic_pillar |
| Helpers | list_timeframes, list_users, list_groups |
Setup
Prerequisites
- Node.js >= 18
- A Perdoo API token (generate one in Perdoo settings)
Install and run
git clone https://github.com/nevilhulspas/perdoo-mcp.git
cd perdoo-mcp
npm install
cp .env.example .env
# Edit .env and add your PERDOO_API_TOKEN
npm run build
npm start
The server starts on port 3001 by default.
Development
npm run dev # Watch mode with tsx
Docker
npm run docker:build
npm run docker:run
Configuration
| Variable | Required | Default | Description |
|---|---|---|---|
PERDOO_API_TOKEN |
Yes | - | Bearer token for the Perdoo GraphQL API |
PORT |
No | 3001 |
HTTP server port |
NODE_ENV |
No | development |
Environment (development or production) |
MCP Client Configuration
Local
{
"mcpServers": {
"perdoo": {
"url": "http://localhost:3001/mcp"
}
}
}
Remote
When deployed to a remote server with a domain:
{
"mcpServers": {
"perdoo": {
"url": "https://perdoo-mcp.yourdomain.com/mcp"
}
}
}
Remote Deployment (Dokploy)
This server is designed to run remotely via its HTTP transport. The included Dockerfile makes it straightforward to deploy on any container platform. Below are instructions for Dokploy on a VPS (e.g. Hetzner).
1. Create the application
In your Dokploy dashboard:
- Create a new Application
- Set source to Git and point it to this repo (
https://github.com/nevilhulspas/perdoo-mcp.git) - Set build type to Dockerfile (it will auto-detect the
Dockerfilein the repo root)
2. Configure environment variables
In the application's Environment tab, add:
PERDOO_API_TOKEN=your-bearer-token-here
PORT=3001
NODE_ENV=production
3. Configure the domain
In the Domains tab:
- Add your domain (e.g.
perdoo-mcp.yourdomain.com) - Enable HTTPS (Dokploy handles Let's Encrypt automatically via Traefik)
- Set the container port to
3001
4. Health check
Configure the health check in Dokploy to use:
- Path:
/health - Port:
3001 - Expected status:
200
5. Deploy
Click Deploy. Dokploy will build the Docker image and start the container.
Security considerations
The MCP endpoint has no built-in authentication. Anyone who can reach your server can use the Perdoo tools with your API token.
For production deployments, secure the endpoint using one of these approaches:
- Traefik BasicAuth middleware - Add basic auth via Dokploy's Traefik config labels
- API key middleware - Add a custom Express middleware that checks an
Authorizationheader (see Auth section below) - IP allowlisting - Restrict access to known IPs via firewall rules or Traefik's
ipAllowListmiddleware - VPN/Tailscale - Run the server on a private network only accessible via VPN
Session persistence
Sessions are stored in-memory. A container restart (deploy, crash, scaling) drops all active MCP sessions. Clients will need to re-initialize. This is expected behavior for MCP's Streamable HTTP transport.
Resource sizing
The server is lightweight (single-concurrency request queue, no database). A minimal VPS (1 vCPU, 512 MB RAM) is sufficient.
Endpoints
| Method | Path | Description |
|---|---|---|
POST |
/mcp |
MCP request handler (creates sessions on initialize) |
GET |
/mcp |
SSE endpoint for server notifications |
DELETE |
/mcp |
Session termination |
GET |
/health |
Health check (status, version, active sessions) |
Architecture
src/
├── server.ts # Express HTTP server with session management
├── lib/
│ ├── env.ts # Environment validation (Zod)
│ ├── errors.ts # Typed error classes
│ └── logger.ts # Structured logging
├── mcp/
│ ├── index.ts # MCP server factory
│ └── tools/ # Tool definitions (one file per entity)
└── services/perdoo/
├── client.ts # GraphQL client with resilience stack
├── types.ts # TypeScript types and enums
├── operations/ # GraphQL queries and mutations
├── circuit-breaker.ts # Fails-open after 5 failures
├── rate-limiter.ts # Token bucket (30 capacity, 3/sec)
├── request-queue.ts # Single concurrent request
└── retry.ts # Exponential backoff (queries only)
Resilience
The client applies a resilience stack in order:
- Request Queue - Ensures only one request at a time
- Circuit Breaker - Opens after 5 consecutive failures, resets after 30s
- Retry - Exponential backoff for queries only (mutations are never retried)
- Rate Limiter - Token bucket with 30 capacity, refilling at 3 tokens/sec
Adding Authentication
The server does not include authentication out of the box. For production use, add a shared-secret middleware:
- Add a
MCP_API_KEYenvironment variable - Add middleware to
src/server.tsthat checks theAuthorization: Bearer <key>header on/mcproutes - Configure your MCP client to send the header:
{
"mcpServers": {
"perdoo": {
"url": "https://perdoo-mcp.yourdomain.com/mcp",
"headers": {
"Authorization": "Bearer your-mcp-api-key"
}
}
}
}
Alternatively, use Traefik's built-in BasicAuth middleware at the reverse proxy level (no code changes required).
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。