business-days-mcp
A server for business day arithmetic and SLA calculations, supporting holidays for ~150 countries, plus cron and recurrence-rule parsing.
README
business-days-mcp
An MCP server for the date arithmetic that support, billing and operations teams actually do: is this a working day, when is this SLA due, what is the next working day after a bank holiday — plus cron and recurrence-rule parsing for the recurring side of the same job.
It runs offline. No network calls, no stored data, no model calls — every answer is computed from bundled public-holiday rules.
What it does
Business days and SLAs — holiday-aware for ~150 countries and their
subdivisions (England and Scotland differ; so do US states), via the
holidays package.
| Tool | Answers |
|---|---|
is_business_day |
Is this a working day? If not, is it a weekend or a named holiday? |
next_business_day / previous_business_day |
The next/last working day around a date |
add_business_days |
Move ±N working days from a date |
business_days_between |
How many working days between two dates (half-open) |
first_business_day_of_month / last_business_day_of_month |
Month boundaries |
sla_due_time |
When does an N-business-hour SLA fall due, counting only open hours? |
Cron and recurrence — parsing only. Nothing is scheduled, stored, or executed; there are no timers and no job store. If you want something run on a schedule, this is deliberately not that tool.
| Tool | Answers |
|---|---|
validate_cron_expression / explain_cron_expression |
Is this cron valid, and what does it mean in English? |
next_cron_occurrences |
When would it next fire? |
validate_recurrence_rule / explain_recurrence_rule |
Same, for an RFC 5545 RRULE |
next_recurrence_occurrences |
Expand an RRULE to its next occurrences |
Install and run
The hosted endpoint needs no install at all — see
Client configuration. To run it yourself, uvx
fetches and runs it:
uvx --from git+https://github.com/Vessark/business-days-mcp business-days-mcp
From a clone:
git clone https://github.com/Vessark/business-days-mcp
cd business-days-mcp
uv sync && uv run business-days-mcp
uv run pytest # the tests are the documentation
Client configuration
Hosted — nothing to install
The quickest way in. The same tools, no install, no credentials:
{
"mcpServers": {
"vessark": {
"url": "https://vessark.com/mcp"
}
}
}
Run it yourself
If you would rather not depend on someone else's uptime, run it locally — it is the same code, and it never touches the network:
{
"mcpServers": {
"business-days": {
"command": "uvx",
"args": [
"--from",
"git+https://github.com/Vessark/business-days-mcp",
"business-days-mcp"
]
}
}
}
Examples
// is_business_day — 2026-12-28, UK
{"date": "2026-12-28", "country": "GB"}
→ {"is_business_day": false, "reason": "holiday: Boxing Day (substitute day)"}
// sla_due_time — 8 business hours from Monday 16:00
{"start": "2026-07-27T16:00", "business_hours": 8}
→ {"due": "2026-07-28T16:00:00", "started_within_business_hours": true}
// explain_cron_expression
{"expression": "30 9 * * MON-FRI"}
→ {"explanation": "At 09:30, on Monday through Friday."}
Design notes
Every tool is a pure function of its arguments, which is why the test suite is mostly tables. Invalid input is an ordinary typed result, never an exception — a model that gets an argument wrong gets a readable reason and can correct itself. Every loop that a caller can influence is bounded, so no request can buy an expensive computation. Error messages never echo what you sent, so the server can't be used to relay text into someone else's agent.
Known limits, stated rather than hidden: a working week is Monday–Friday, so countries with Sunday–Thursday weeks are wrong under this assumption; holiday data covers only the years the underlying dataset knows about, and a date outside that range is refused rather than silently answered as an ordinary working day.
Prior art
fbdo/business-day-mcp covers
business-day and holiday arithmetic and predates this server. If that is all
you need, use it — it also exposes timezone-aware "today" and holiday listing,
which this server does not.
This one was built for a different centre of gravity: SLA clocks that count
only business hours, and cron / RRULE parsing for the recurring side of the
same work. The overlap in the business-day tools is real and the names are
conventional; both servers are MIT licensed and both build on the excellent
holidays package.
Releasing
Bump the version in both pyproject.toml and server.json, then push a
matching tag:
git tag v0.2.0 && git push origin v0.2.0
That runs the test suite, checks the tag agrees with both version fields, and republishes to the MCP registry. Authentication is GitHub OIDC, so no token is stored anywhere.
About
Built and maintained by Vessark. The tools in this repository are the free tier and are open source under the MIT licence. Vessark's workspace tools — the ones that operate on a connected workspace — are the commercial product and are not open source.
Issues and pull requests welcome.
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。