Launching today
Vantaj Uptime New Features (MCP & more)
Uptime monitoring your AI agent can talk to
21 followers
Uptime monitoring your AI agent can talk to
21 followers
Vantaj is uptime, SSL, domain, and heartbeat monitoring with zero false positives. This week we're launching an MCP server (talk to your monitors from Claude or Cursor), one-click migration from UptimeRobot/Pingdom/StatusCake/Better Stack/Kuma, and inverted checks for exposure monitoring.
How does the MCP server handle auth for letting an LLM talk to monitors, and can you scope what data the model can actually see and trigger?
@serkangzelytiq Important questions!
The MCP server authenticates with the same API key as the REST API. The server is stateless Streamable HTTP, so no sessions or anything, just the API key as a bearer header. Same key works for both REST and MCP, so nothing new to manage.
Since we're "riding" on our already granular API keys, you can scope per resource type (read-only or read+write on monitors, heartbeats, domains, status pages, incidents...).
For guardrails there's a structural limit baked into the MCP surface itself, it's reads plus additive writes only, so delete tools and billing settings aren't exposed at all, with anything destructive staying in the dashboard and REST API where a human is in the loop. (That's true even for a key with full write scope.)
The MCP server works surprisingly well in Claude, way more useful than I expected for quick checks on my heartbeats. Migration from UptimeRobot was painless too, took maybe two minutes total.
Finally checked out the MCP integration and it actually works as advertised. I ran a quick check on a couple of monitors through Claude and the response was snappy, no fuss.