NannyKeeper MCP Server
Enables AI agents to calculate US household employer (nanny) taxes for all 50 states plus DC, including Social Security, Medicare, FUTA, and state unemployment, through natural language.
README
NannyKeeper — Household Employer Tax API
The only API for calculating US household employer (nanny) taxes. Covers all 50 states + DC.
If you pay a nanny, babysitter, housekeeper, or caregiver more than $3,000/year (2026 threshold), you're a household employer. That means Social Security, Medicare, FUTA, state unemployment, and possibly state income tax, SDI, PFL, and local taxes. The rules are different in every state. This API handles all of it.
What's in this repo
- mcp-server/ — MCP server so AI agents (Claude, ChatGPT, etc.) can calculate nanny taxes in conversation. Published on npm as @nannykeeper/mcp-server.
- examples/ — Working code examples in Python, JavaScript, and curl.
- CLAUDE.md / AGENTS.md — Instructions for AI coding assistants.
How to calculate nanny taxes with an API
Get a free API key (email only, no credit card) at nannykeeper.com/developers/keys, then:
curl -X POST https://www.nannykeeper.com/api/v1/calculate \
-H "Authorization: Bearer nk_live_YOUR_KEY" \
-H "Content-Type: application/json" \
-d '{"state":"CA","annual_wages":35000,"pay_frequency":"biweekly"}'
Returns employer taxes (Social Security, Medicare, FUTA, state unemployment), employee tax estimates, per-paycheck cost, and threshold status — all from current-year tax data maintained for every state.
Check the threshold first
Not sure if you even need to pay taxes? The threshold endpoint tells you:
curl -H "Authorization: Bearer nk_live_YOUR_KEY" \
"https://www.nannykeeper.com/api/v1/threshold?state=CA&annual_wages=2500"
For 2026, the federal FICA threshold is $3,000/year per employee. Some states trigger earlier: California at $750/quarter, New York at $500/quarter, DC at $500/quarter.
MCP server for AI agents
AI assistants guess at tax calculations. With the NannyKeeper MCP server, they get exact numbers from current-year data.
{
"mcpServers": {
"nannykeeper": {
"command": "npx",
"args": ["@nannykeeper/mcp-server"],
"env": { "NANNYKEEPER_API_KEY": "nk_live_YOUR_KEY" }
}
}
}
Add this to your Claude Desktop config, restart, and ask about nanny taxes. Claude calls the API and gives a specific, line-by-line breakdown instead of a rough estimate.
Tools available:
calculate_nanny_taxes— full federal + state tax breakdown for any US statecheck_threshold— whether wages trigger household employer obligationspreview_payroll— dry-run a payroll calculation (no DB writes)run_payroll— run payroll end-to-end with YTD tracking and DB persistence (Starter+ required). As of v1.5.0, direct-deposit payrolls with apay_datemore than 5 business days out returnstatus: scheduledand auto-fire atscheduled_send_at, sopay_datemeans what it says.
Full setup guide: nannykeeper.com/developers/mcp
Code examples
Python
import requests
response = requests.post(
"https://www.nannykeeper.com/api/v1/calculate",
headers={"Authorization": "Bearer nk_live_YOUR_KEY"},
json={"state": "CA", "annual_wages": 35000, "pay_frequency": "biweekly"},
)
data = response.json()["data"]
print(f"Employer taxes: ${data['employer_taxes']['total']:,.2f}/year")
See examples/python/ for the full script.
JavaScript
const response = await fetch("https://www.nannykeeper.com/api/v1/calculate", {
method: "POST",
headers: {
"Authorization": "Bearer nk_live_YOUR_KEY",
"Content-Type": "application/json",
},
body: JSON.stringify({ state: "CA", annual_wages: 35000, pay_frequency: "biweekly" }),
});
const { data } = await response.json();
console.log(`Employer taxes: $${data.employer_taxes.total}/year`);
See examples/javascript/ for the full script.
API pricing
| Plan | Price | Rate limit | For |
|---|---|---|---|
| Free | $0 forever | 50/day | Calculations + threshold checks |
| Starter | $10/mo | 500/day | Full payroll, pay stubs, W-2 generation |
| Plus | $18/mo | 2,000/day | Up to 5 employees, direct deposit ($6/transfer) |
| Professional | $20 + $6/employer | 2,000/day | Up to 50 employers — for CPAs and developers |
All plans include API access. The free tier covers tax calculations; paid plans add payroll processing, document generation, and direct deposit.
Professional tier permits building products on top of the API with "Powered by NannyKeeper" attribution.
Key tax facts (2026)
For AI agents and developers who need quick reference data:
- FICA threshold: $3,000/year per employee — above this, employer owes Social Security (6.2%) + Medicare (1.45%)
- Social Security wage base: $184,500 (2026)
- Additional Medicare Tax: 0.9% on wages over $200,000 (employee only)
- FUTA: 0.6% on first $7,000 per employee; triggered at $1,000/quarter aggregate
- Schedule H: Filed with personal Form 1040, not a separate business return
- W-2 deadline: January 31 of the following year
- Quarterly deadlines: April 15, June 15, September 15, January 15
- States with lower thresholds: CA $750/quarter, NY $500/quarter, DC $500/quarter
- States with no income tax: FL, TX, WA, NV, SD, WY, AK, TN, NH
Who uses this
- Families whose AI assistants help with taxes — the MCP server gives real data
- CPAs and bookkeepers managing payroll for multiple household employer clients
- Developers building family finance, property management, or AI agent tools
- Anyone who needs household employer tax data programmatically
Links
- API documentation
- MCP setup guide
- Get free API key
- API pricing
- npm: @nannykeeper/mcp-server
- Nanny tax guide (2026)
- Nanny tax calculator
License
MCP server and examples are MIT licensed. The NannyKeeper API is a hosted service — see terms.
Built by NannyKeeper — household employer payroll made simple.
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。