Monitor Your Websites from Claude Desktop: Baromio Blog

TL;DR

Connect Claude to Baromio in a couple of minutes and ask about your monitors, uptime and incidents in plain language. Reading works on every plan; creating or changing monitors from Claude needs Pro or Business.

If you already use Claude, you can skip the dashboard for routine questions about your sites. Connect it once, then ask which monitors are down, how uptime looked this month, or what happened during last night's incident.

Setup takes a couple of minutes and a Baromio account. The Free plan is enough for steps 1 and 3.

What you'll need

A Baromio account. The Free plan covers reading monitors, uptime and incidents.
Claude, either the desktop app or claude.ai. Claude's Free plan allows one custom connector.

Step 1: Connect Claude to Baromio

In Claude, go to Customize → Connectors.
Click + → Add custom connector and paste the Baromio MCP server URL:
```
https://baromio.io/mcp
```
Claude opens a browser window. Log in to Baromio and click Authorize.
Back in Claude, ask: List my Baromio monitors. On a new account the list is empty. That is expected, and it confirms the connection works.

No API key and no config file: Claude handles the OAuth sign-in for you.

Step 2: Create a monitor by describing it (Pro and Business)

Creating, pausing or changing monitors from Claude needs a Pro or Business plan and a connection authorized with full access. On the Free plan Claude gets an upgrade message instead, and you create monitors in the dashboard.

With write access, a request like this is enough:

You: Monitor https://example.com every minute and email me when it goes down.

Claude: [calls create-monitor] Done. I created an HTTP monitor for https://example.com that checks every minute.

Claude picks the monitor type and interval from your request and creates the monitor through the same API the dashboard uses. The shortest interval depends on your plan: 5 minutes on Free, 1 minute on Pro and 30 seconds on Business.

Step 3: Ask about status in plain language

Once monitors are running, ask instead of opening the dashboard. This works on every plan:

You: What was my uptime over the last 30 days?

Claude: [calls get-uptime-stats] example.com was up 99.95% of the time over the last 30 days, with an average response time of 312 ms.

You: What happened in the last incident?

Claude: [calls get-monitor-incidents] The latest incident started on 12 September at 03:14 and lasted 4 minutes. The recorded cause is a connection timeout.

Uptime statistics cover the last 24 hours or the last 30 days. On Pro and Business, Claude can also call diagnose-incident to write an AI Incident Diagnosis that explains the likely cause in plain language.

What else write access unlocks

On Pro or Business, Claude can also pause and resume monitors, change check intervals and alert thresholds, and delete monitors. Deleting needs the exact monitor ID, so a vague request cannot remove the wrong one. That helps during maintenance: "pause the staging monitor" instead of finding it in a list.

Full reference

For the complete tool list, token types and troubleshooting, see the AI Agents & MCP documentation.

The conversations above are illustrative. The tool names and the kind of answer are real; the monitor, numbers and dates are made up.