What is Uptime Kuma?
Uptime Kuma checks whether your services are up and alerts you when they are not. It supports HTTP(s), TCP, keyword, JSON query, WebSocket, ping, DNS, push and Docker container monitors, checks as often as every 20 seconds, and sends alerts to Telegram, Discord, Slack, email and, by the README's count, more than 90 notification services. It also publishes status pages, which can be mapped to your own domains.
It has no AI features of its own. It earns a place here because it is the watchdog most self-hosted AI stacks run: Open WebUI's own monitoring guide is written around it, with a /health check, a JSON query that confirms Ollama models are reachable, and a deeper check that sends a real chat completion.
The project is MIT-licensed, written in JavaScript and runs from one Docker container on port 3001. There is no official hosted version or paid tier; the README warns that the top Google result, uptimekuma.co, is not the project's site.
Is Uptime Kuma worth it?
Uptime Kuma answers one question well: is the thing still up? With 94,486 GitHub stars and version 2.5.11 released on 12 November 2026, it is the default choice for people who host their own services, and that now includes local model servers and chat front ends. Open WebUI's monitoring documentation uses it for three levels of check: a /health ping every 60 seconds, a JSON query on /api/models every 300 seconds that can count Ollama models specifically, and a check that sends a real chat completion and validates the reply.
That last check is the useful one for AI work. A model server can return 200 on its health route while the model itself fails to load; a JSON-query or keyword monitor that posts a prompt catches that. The trade-off is that the monitor then holds an API key in its settings, alongside the notification tokens for Slack, Telegram and the rest, so the Uptime Kuma instance itself becomes something worth protecting.
That matters because its security record is busy. GitHub lists a high-severity unauthenticated denial-of-service advisory published on 9 October 2026 and an authenticated remote code execution flaw through Apprise argument injection published on 2 November 2026, and 2.5.11 fixed an admin-only remote code execution caused by an upstream package. Notification templates had server-side template injection bugs that allowed file reads in 2025 and again in March 2026. The default Docker command listens on all interfaces; the README shows how to bind to 127.0.0.1 instead.
Uptime Kuma is not LLM observability. It does not see tokens, latency per model, cost or answer quality; Langfuse or PostHog's AI observability do that job. Version 2.0.0, released on 20 October 2025, carried breaking changes and a migration guide, so upgrades across major versions need a backup first. Network file systems such as NFS are not supported for its data directory.
Where it earns its place
A homelab owner running Ollama and Open WebUI on one GPU box
Find out when the chat front end or the models go down, before the family does
Add an HTTP monitor on Open WebUI's /health route and a JSON-query monitor on /api/models that counts Ollama models, both alerting to Telegram. It costs nothing beyond the container, and the checks follow Open WebUI's own guide.
A two-person agency hosting client chatbots on a VPS
Give clients a status page and get paged when a bot stops answering
One status page per client, mapped to the client's own domain, with a monitor that posts a test prompt and checks the reply for an expected keyword. Alerts go to Slack; there is no per-monitor or per-seat fee.
An ML engineer with a nightly batch job that calls a model API
Know when the job silently stops running
A push monitor waits for the job to call its URL at the end of each run and alerts if the call does not arrive on schedule. It needs no inbound access to the batch server.
Before you commit
- Setup
- docker compose up -d with the project's compose.yaml, or one docker run command; it serves on port 3001. Non-Docker installs need Node.js 20.4 or later and pm2, and FreeBSD is not supported.
- Integrations
- HTTP(s), TCP, keyword, JSON query, WebSocket, ping, DNS, push and Docker container monitors; more than 90 notification services including Telegram, Discord, Slack, Pushover and SMTP email.
- Data and privacy
- Everything stays on your server. The instance stores notification tokens and any headers or API keys you give monitors, and listens on all interfaces by default.
- Cost at scale
- No licence or hosted fee at any size. Your only cost is the server it runs on; resource figures for large monitor counts are not published.
- Limits
- 850 open issues and pull requests, and the README says the maintainer does not have time to test every pull request. Major upgrades, such as 1.x to 2.0, need the migration guide and a backup.
What’s good
- One Docker container, running in minutes, no external database required
- Monitors that can post a prompt and check the reply, not just ping a port
- More than 90 notification services and public status pages
- MIT licence with no paid edition holding features back
What’s not
- Several high-severity security advisories in October and November 2026
- Stores API keys and notification tokens, so the instance needs protecting
- No token, cost or quality metrics for model calls
- Data directory cannot sit on NFS
Best for
Self-hosters and small teams running Ollama, Open WebUI or other AI services who want an alert the moment an endpoint or model stops responding.
Uptime Kuma: common questions
- Is Uptime Kuma worth it?
- The watchdog to put next to a self-hosted model server: quick to run, free, and it will page you when Ollama dies, but it measures availability, not model quality. We rate it 4.2 out of 5.
- How much does Uptime Kuma cost?
- Free and open source under MIT. No official hosted or paid tier; you run it yourself.
- What are the downsides of Uptime Kuma?
- Several high-severity security advisories in October and November 2026; stores API keys and notification tokens, so the instance needs protecting; no token, cost or quality metrics for model calls; data directory cannot sit on NFS.
- Who is Uptime Kuma best for?
- Self-hosters and small teams running Ollama, Open WebUI or other AI services who want an alert the moment an endpoint or model stops responding.
- How much does Uptime Kuma cost for a team or in production?
- No licence or hosted fee at any size. Your only cost is the server it runs on; resource figures for large monitor counts are not published.
- Is Uptime Kuma safe for company data?
- Everything stays on your server. The instance stores notification tokens and any headers or API keys you give monitors, and listens on all interfaces by default.
Answers checked Nov 16, 2026.