bamboohr-mcp
A read-only MCP server for BambooHR that enables safe AI assistant access to employee records, time-off, files, and directories via natural language queries.
README
bamboohr-mcp
A read-only Model Context Protocol server for BambooHR.
This is a fork of evrimalacan/mcp-bamboohr (MIT). See NOTICE for attribution and the full list of changes.
Why this fork
BambooHR is sensitive HR data, so this fork hardens the upstream server for exposing it to an AI assistant safely. Design priorities: read-only by construction, least privilege, and the host process holds no credential. The upstream server is a clean, well-tested base; this fork hardens it for that posture:
- Read-only by construction. The HTTP client exposes only
get/getBuffer—post/put/deletewere removed, so no write tool can be added by accident. Every registered MCP tool is aGET. - Proxy-ready, credential-free. Set
BAMBOO_BASE_URLto a loopback token-proxy and the server sends noAuthorizationheader — the proxy injects a fresh OAuth Bearer per request. The BambooHR credential never lives in this process. - Injectable config. The credential is resolved in
config.tsand passed into the client, not hardwired into a boot-time singleton — so a future multi-tenant/remote host can construct a per-request client without touching the tool layer.
Configuration
Two shapes (see .env.example):
Proxied (recommended for shared / sensitive use):
{
"mcpServers": {
"bamboohr": {
"command": "node",
"args": ["build/index.js"],
"type": "stdio",
"env": {
// A loopback proxy injects Authorization; no token here.
"BAMBOO_BASE_URL": "http://127.0.0.1:7339"
}
}
}
}
Direct (a local/desktop setup with your own API key):
{
"mcpServers": {
"bamboohr": {
"command": "node",
"args": ["build/index.js"],
"type": "stdio",
"env": {
"BAMBOO_API_TOKEN": "your_api_token",
"BAMBOO_COMPANY_DOMAIN": "your_company_subdomain"
}
}
}
}
Tools (all read-only)
| Tool | Purpose |
|---|---|
get-employee |
Employee record with selectable fields |
get-employee-photo |
Employee photo by size |
get-employee-directory |
Company-wide directory |
get-employee-goals |
Performance goals for an employee |
estimate-time-off-balance |
Projected time-off balances |
get-time-off-requests |
Time-off requests (filterable) |
get-whos-out |
Upcoming time off + holidays |
list-company-files |
Browse company files/categories (metadata) |
get-company-file |
Download a company document by id |
get-meta-fields |
Discover available BambooHR data fields |
Egress note. Some tools (directory, file download) can return large volumes of PII. For shared / sensitive deployments, enforce row/size caps and field scoping at the proxy in front of this server, rather than trusting the tool layer.
Development
npm install
npm run build
npm test
npm run dev # watch mode
TypeScript (strict), ESM, Node ≥22. Tests run under Jest.
Scope
Read-only data access only. There are deliberately no write tools. Adding any write capability is an explicit decision that changes the security posture — not a casual PR.
License
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。