@byterover/umami-mcp
A read-only MCP server for Umami analytics, enabling natural language queries of website stats, traffic trends, events, sessions, and analytics reports.
README
@byterover/umami-mcp
A minimal, read-only Model Context Protocol server for Umami analytics — Cloud or self-hosted.
It exposes a small set of read tools (list sites, stats, time series, top metrics, live visitors) over stdio, so an MCP client such as Grove can let an agent answer questions about your web analytics. It issues no writes — there are no create/update/ delete tools, by design.
Why this exists
Community Umami MCP servers exist but have little usage and aren't reviewed by anyone we trust with an analytics credential. This is byterover's first-party, source-available wrapper: small enough to read end-to-end, read-only, and published with provenance. We dogfood it on our own landing-page analytics.
Install
npx @byterover/umami-mcp
Configure
Pick one mode via environment variables.
Umami Cloud — create a read-only API key at cloud.umami.is (Settings → API keys):
UMAMI_API_KEY=your_api_key
Self-hosted — point at your instance and provide a login:
UMAMI_API_URL=https://umami.example.com
UMAMI_USERNAME=your_username
UMAMI_PASSWORD=your_password
Advanced: UMAMI_API_URL overrides the base host (Cloud default
https://api.umami.is) and UMAMI_API_PATH overrides the path prefix (Cloud
/v1, self-hosted /api).
Use with Grove
Add it to your mcp.json as a stdio server (pin the version; keep the key in
.env via a ${VAR} ref):
{
"mcpServers": {
"umami": {
"command": "npx",
"args": ["-y", "@byterover/umami-mcp@0.1.0"],
"env": { "UMAMI_API_KEY": "${UMAMI_API_KEY}" }
}
}
}
Tools surface in Grove as umami__list_websites, umami__website_stats, etc.
Tools
Thirteen read tools covering essentially all of Umami's analytics reads —
consolidated (one metrics tool spans ~10 dimensions; explore_event_data
folds five endpoints behind a mode), never mirroring the REST API 1:1.
Discovery
| Tool | What it returns |
|---|---|
list_websites |
Websites (id, name, domain) these credentials can see. Start here. |
data_range |
Earliest/latest timestamps with data — call before querying ranges. |
Traffic & trends
| Tool | What it returns |
|---|---|
website_stats |
Pageviews, visitors, visits, bounces, total time (with prior period). |
pageviews_series |
Pageviews/sessions time series, bucketed by hour/day/month/year. |
realtime |
Live activity in the last ~30 min (active visitors, recent views/events). |
Breakdowns & events
| Tool | What it returns |
|---|---|
metrics |
Top values for one dimension (url, referrer, browser, country, event, …); expanded=true adds engagement. |
events_series |
Custom-event time series over a range. |
explore_event_data |
Drill into event properties/values (mode: events / properties / fields / stats / values). |
Sessions & journeys
| Tool | What it returns |
|---|---|
list_sessions |
Individual visitor sessions (paginated, searchable). |
session_detail |
One session's summary + activity log + custom properties. |
Analyses (compute-reads, POST — still read-only)
| Tool | What it returns |
|---|---|
funnel_report |
Conversion funnel across ordered steps (paths/events). |
retention_report |
Return-visitor retention over the range (needs a timezone). |
journey_report |
Common navigation paths between a start and (optional) end step. |
Range tools accept ISO startAt/endAt; report tools accept ISO
startDate/endDate (e.g. 2026-07-01). Omit them for the last 7 days.
Develop
pnpm install
pnpm build # tsc → dist/
pnpm typecheck
pnpm test # keyless, network-free
pnpm lint
License
Elastic License 2.0 — © byterover.
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。