Free vs paid monitoring MCP servers observability already has a bill.

Whether your monitoring MCP setup costs anything depends almost entirely on the layer below. Netdata and SigNoz can be fully self-hosted open source — bridge and backend both free, infrastructure costs aside. sentry-mcp and Dynatrace-mcp connect to commercial services where your existing plan is the cost, and the MCP layer adds nothing to the bill.

Hobbyist-scale options round it out: HomeLab Monitor bundles a read-only MCP server into a self-hosted dashboard, which is about as free as observability gets.

Monitoring is the worst system to have go down silently, and self-hosting it means you're the SLA. Storage for metrics retention, alert routing, upgrade discipline — running SigNoz or Netdata well is a genuine ops responsibility, which is exactly why commercial observability vendors exist.

The MCP angle adds one more consideration: an agent reading your logs and metrics sees everything in them. Self-hosting keeps that data in-house; SaaS bridges extend whatever trust you've already placed in the vendor.

The interesting paid frontier in monitoring isn't dashboards — it's discrete checks. A Lighthouse performance run, the kind lighthouse-mcp performs, is a self-contained unit of work with a clear result. So is an uptime probe from a specific region, an SSL certificate validation, or a one-off load test. These don't require an ongoing relationship; they're verbs, not subscriptions.

That's the shape paid hosted endpoints can serve well: an agent that needs to verify a deploy from outside its own network pays for exactly one external check.

With x402, a check endpoint replies HTTP 402 with a price, the agent pays in USDC on Base — about two seconds to settle — and the probe runs only after payment clears. From $0.01 per check, with an Ed25519-signed receipt as proof the check happened and what it cost. Signed receipts are unusually useful in monitoring, where 'we verified this at 14:02' is sometimes the whole point.

Per-check pricing fits agent-driven operations cleanly: a deploy pipeline that pays a cent to confirm the site is up beats a monthly synthetic-monitoring plan it pings twice a week.

Bridge your existing stack with the free connector built for it — that decision was made when you chose Netdata, Sentry, or SigNoz. Buy per-call where the unit is a discrete check from infrastructure you don't run. Loomal's Monitoring category puts all 114 live listings in one view, with x402 pricing visible wherever a maintainer has set it.