Home
Monitoraggio dell'uptime Observability Real User Monitoring Risposta agli incidenti Integrazioni & API Esplora la piattaforma →
Prezzi
Calcolatore di uptime Cron Expression Builder Codici di stato HTTP Tutti gli strumenti gratuiti
Piattaforma Blog Chi siamo Contatti FAQ
Accedi Inizia gratis
MCP servers compared

Monitoring tools with an MCP server (2026)

Yes, many. As of October 2026 this guide lists 21 vendors besides Site Qwality that publish their own MCP server for uptime, status pages, incident management or observability, so Claude, Cursor, Codex and other assistants can read and change your monitoring. Fifteen are compared side by side from their own docs (hosted or local, sign-in, read-only options, tool count and plans) and six broader platforms are summarized.

Inizia gratis Parla con noi
Every claim linked to the vendor Checked October 17, 2026 Says where we are weaker
Verdetto

Most monitoring vendors now have one. Site Qwality’s is hosted at mcp.siteqwality.com/mcp, has 64 tools (26 only read) and is free on every plan. It is not the biggest: Datadog, Rootly, Better Stack and Cronitor document more tools, Checkly, UptimeRobot, Pulsetic and Uptime.com can create port, TCP or ICMP checks that Site Qwality does not have, and Cronitor, Sentry, Grafana, Uptime.com and Rootly also offer a local server, which ours does not. Choose by what you want to ask your assistant, then by how tightly you can restrict it.

Ultima verifica il sulle pagine ufficiali di ciascun fornitore. Vedi le fonti.

In sintesi

  • Fifteen vendors besides Site Qwality are in the table: Better Stack, UptimeRobot, Checkly, Cronitor, OpenStatus, Pulsetic, HetrixTools, Hyperping, Uptime.com, Datadog, Sentry, Grafana, PagerDuty, incident.io and Rootly. New Relic, Dynatrace, Honeycomb, Elastic, Splunk and Atlassian are summarized below it. We found no official server from StatusCake, Pingdom, Uptime Kuma, Instatus or Healthchecks.io.
  • All sixteen servers in the table have write tools, so the question is how you restrict them. UptimeRobot, OpenStatus, Hyperping, HetrixTools, Grafana and Site Qwality offer a read-only choice at sign-in or on the key, and Rootly has a switch that turns write tools off. Better Stack uses tool allowlist headers. Pulsetic and Uptime.com document no read-only mode.
  • Plans: Site Qwality includes it on every plan, HetrixTools on all plans including Free, and Checkly shows it on all four. Hyperping and incident.io do not include it on their free tiers. Most other vendors’ MCP docs do not say which plans include it.
  • Largest tool surfaces: Datadog’s docs list 383 tools and Better Stack’s more than 100 (our counts), Rootly says 200+ and Cronitor says 71. Site Qwality has 64.
  • Site Qwality gaps: hosted only, no port, ping or TCP checks, and only HTTP monitors can be created with its curated tools.

Sixteen MCP servers side by side

Vendor Hosted or local Sign-in Read-only option Tools Plans
Site Qwality Hosted only OAuth or API key Yes, at sign-in or by key scope 64 (26 read only) Free on every plan
Better Stack Hosted OAuth or API token Tool allowlist and blocklist headers 100+ listed Not stated in its MCP docs
UptimeRobot Hosted OAuth or API key Yes, read-only key or OAuth choice 32 Not stated in its MCP docs
Checkly Hosted OAuth (approved clients) or API key By permission, no read-only switch 49 listed MCP row ticked on all four plans
Cronitor Hosted, plus a local CLI server OAuth or API key By user role 71 Not stated in its MCP docs
OpenStatus Hosted OAuth or API key Yes, read-only scope hides write tools 29, plus 10 incident tools No plan limit stated
Pulsetic Hosted API key (OAuth in ChatGPT) None documented 17 Docs and pricing page differ
HetrixTools Hosted API key Yes, by key scope Not stated All plans, including Free
Hyperping Hosted API key, one project per key Yes, read-only key 49 (28 read, 21 write) Not on Free; on Essentials, Pro and Business
Uptime.com Hosted, plus a local server API token or OAuth2 None documented Not stated Not stated in its MCP docs
Datadog Hosted OAuth, tokens or keys Org setting hides write tools 383 listed Not stated in its MCP docs
Sentry Hosted and local OAuth (hosted), token (local) Yes, by limiting skills Not stated Not stated in its MCP docs
Grafana Hosted (Cloud) and local (open source) OAuth 2.1 (Cloud), service account token (local) Yes, on both Not stated Cloud server is part of Grafana Assistant
PagerDuty Hosted (local server deprecated) API key or OAuth Via a scoped OAuth client 18 listed Not stated in its MCP docs
incident.io Hosted OAuth or API key None documented; API keys take chosen scopes Not stated Team, Pro and Enterprise; not Basic (free)
Rootly Hosted, plus a local server OAuth2 or API token Yes, a switch that turns write tools off 200+ Not stated in its MCP docs

Tools is the vendor’s own figure where it gives one. “Listed” is our count of the tools on its docs page. Checked October 17, 2026. Sources below.

Inizia gratis

Uptime, status page and cron monitoring vendors

Ten vendors in this group publish a server you add by URL. Everything about the nine competitors below comes from their own docs, checked October 17, 2026.

  • Site Qwality. Hosted at https://mcp.siteqwality.com/mcp, with OAuth sign-in or an API key. 64 tools: 26 only read and 38 can change things. They cover monitors, incidents, status pages and read-only log, metric and trace search, and the rest of the REST API through two generic tools. Free on every plan. Details are on the MCP server page.
  • Better Stack. Hosted at https://mcp.betterstack.com. OAuth is recommended, or you pass an API token. There is no read-only mode, but the headers X-MCP-Tools-Only and X-MCP-Tools-Except allowlist or blocklist tools, and the docs suggest that for read-only access. They list more than 100 operations (our count) across uptime, incidents, on-call, status pages, logs, metrics and errors, and do not say which plans include it.
  • UptimeRobot. Hosted at https://mcp.uptimerobot.com/mcp. Sign in with OAuth and approve read-only or full access, or use an API key (a read-only key works). The repository says 32 tools. It can create HTTP, keyword, ping, port, heartbeat and DNS monitors, and its guide says “Delete operations are not exposed over MCP.” MCP calls count against the same rate limit as the API.
  • Checkly. Hosted at https://api.checklyhq.com/mcp. OAuth works for Checkly-approved clients (dynamic client registration is rejected), or use an API key; service keys need a Starter, Team or Enterprise plan. Tools appear only if the session and your account role allow them, and there is no separate read-only flag. The tools page lists 49 tools (35 read, 14 write), including create-check for URL, TCP, DNS, ICMP, traceroute, gRPC and SSL monitors, and AI root cause analysis. Its pricing page ticks MCP on all four plans.
  • Cronitor. Hosted at https://cronitor.io/mcp, plus a separate local server built into CronitorCLI that talks over stdio. OAuth or an SDK Integration API key. The docs say “71 tools in total” and “Read-only users can only call read tools,” so access follows the user’s role. It covers cron, heartbeat and uptime monitors, status pages, issues and maintenance windows. Its agent quick start says individual run logs are not available over MCP. Plans are not stated.
  • OpenStatus. Hosted at https://api.openstatus.dev/mcp. OAuth 2.1, or an API key as a fallback. You pick a read-only scope at consent or when creating the key. In that scope, the reference says mutation tools “are not advertised in tools/list and cannot be called.” It exposes 29 tools, 27 on plans without the audit-log feature, plus 10 more for workspaces with incident management. Monitors are read only over MCP; status reports, maintenance and incidents are writable. No plan limit is stated.
  • Pulsetic. Hosted at https://api.pulsetic.com/mcp, 17 tools, authenticated with an API key sent as a Bearer token (ChatGPT signs in with OAuth). Its docs say “Your API key has full access to your Pulsetic account” and document no read-only mode. create_monitor makes HTTP, TCP or ICMP monitors. Its pricing page ticks API/MCP/N8N on Solo, Team and Organization and not on Free, while the docs ask for “a plan with API access (Business or Enterprise)”, so confirm the plan with Pulsetic.
  • HetrixTools. Hosted at https://mcp.hetrixtools.net over Streamable HTTP, with an API key. Its MCP page says “MCP access is available on all Uptime Monitor and Blacklist Monitor plans, including our Free for life plans” and suggests giving the agent a read-only key. It reaches uptime monitors, server agents, blacklist monitors, maintenance windows, status pages and the activity log. No tool list or count is published.
  • Hyperping. Hosted at https://api.hyperping.io/v1/mcp over Streamable HTTP, with an API key sent as a Bearer token. Each key belongs to one project and is either read_only or read_write. The docs list 49 tools (28 read, 21 write) and its agent setup page says write tools refuse a read-only key in the server. Write tools cover monitors, outages, status pages, status page incidents and maintenance windows. Agents can pause monitors but not delete them, and its pricing page shows the MCP server as not included on Free and included on Essentials, Pro and Business.
  • Uptime.com. Hosted at https://mcp.uptime.com/mcp, and the open source uptime-mcp server also runs locally over stdio, as a container or from Go. Sign in with an API token or an OAuth2 browser flow. The README documents no read-only mode and no tool count. It can create and update checks of many types, including HTTP, DNS, SSL, ICMP, TCP, UDP, SMTP and SSH, and it has delete_check and create_status_page_incident. Plans are not stated.

Observability and incident platforms

These six are broader products. None of them is a status page or uptime tool first, so they overlap with Site Qwality only in part.

  • Datadog. Hosted, with a regional endpoint per Datadog site, for example https://mcp.datadoghq.com/v1/mcp for US1. It is not GovCloud compatible. OAuth 2.0 is recommended, and tokens or API and application keys work for servers and CI. An org-level MCP Write Access setting hides write tools from every user, and roles also need the mcp_write permission. Without a toolsets parameter only the core tools load, and the tools reference lists 383 (our count). It covers logs, metrics, traces, monitors, incidents and Synthetic HTTP API tests, and the docs call it “under significant development.”
  • Sentry. Hosted at https://mcp.sentry.dev/mcp with OAuth, and a local stdio server, @sentry/mcp-server, that uses a Sentry token. The repository says stdio is still a work in progress. Access is split into skills: inspect is read only, while triage and project-management can resolve issues, create projects and alert rules, and create, update or delete uptime monitors. The repository documents narrowing the exposed tools with a ?skills= query parameter. Plans are not stated.
  • Grafana. Two servers. The open source mcp-grafana runs locally with a service account token and a --disable-write flag. The Grafana Cloud server is hosted at https://mcp.grafana.com/mcp with OAuth 2.1, and read-only means clearing both Write and Query access at consent. It reaches Prometheus, Loki, Tempo and Pyroscope data, dashboards, alerting, Incident and OnCall. The Cloud server is part of Grafana Assistant, and connecting through it counts as active use for that billing.
  • PagerDuty. Hosted at https://mcp.pagerduty.com/mcp (EU accounts use mcp.eu.pagerduty.com), with an API key or OAuth. The docs list 18 tools, each with several actions, covering incidents, services, schedules and on-calls, teams, escalation policies and status pages. Tool filtering is not available on the hosted server, so you limit it with a scoped OAuth client. The self-hosted server is deprecated and its repository archived. The docs do not state plans.
  • incident.io. Hosted at https://mcp.incident.io/mcp, turned on under Settings, MCP. Interactive clients sign in with OAuth and act as your user. For automation you create an API key with the scopes you need and send it as a Bearer token. The docs describe tools for incidents, alerts, escalations, on-call schedules and maintenance windows, including changing who is on call, and say status_page_update only drafts, so a person publishes. They say the server “is available on Team, Pro, and Enterprise plans” and not on Basic (free). Its changelog announced a public beta for paying customers on March 31, 2026. No tool count is published.
  • Rootly. Hosted at https://mcp.rootly.com/mcp, with a Code Mode endpoint and an SSE endpoint, and a local server you start with uvx or Docker. OAuth2 or an API token. The docs say “200+ tools” and offer a hosted slim profile of about 70. The README says setting ROOTLY_MCP_ENABLE_WRITE_TOOLS=false or starting with --no-enable-write-tools restricts the server to read-only tools. It covers incidents, on-call shifts and handoffs, alerts and status pages. Plans are not stated.

Other observability platforms with an official server

These larger suites also publish a server. We read their MCP docs for the basics below and did not put them in the table, because they are not uptime or status page tools first.

  • New Relic. Hosted at https://mcp.newrelic.com/mcp/, with EU and Japan endpoints. OAuth is recommended, or you use a user API key, and some tools work only with OAuth. The overview says “MCP is available to all accounts.” Its setup guide describes an include-tags header that filters tools. Tools include NRQL queries, alerts and issues, and synthetic monitor listings and results, within a limit of 2,000 tool calls per hour per user.
  • Dynatrace. Hosted on your own environment URL, https://{environment-name}.apps.dynatrace.com/platform-reserved/mcp-gateway/v0.1/servers/dynatrace-mcp/mcp. The docs, updated October 6, 2026, say every request needs a bearer token, from a platform token (recommended) or an OAuth client. Tools cover Grail data queries, root cause analysis, Kubernetes and security posture. The earlier open source local server is deprecated.
  • Honeycomb. Hosted at https://mcp.honeycomb.io/mcp, with an EU endpoint. Sign in with OAuth, or use a Management API key for unattended use. The configuration guide says write tools, such as creating boards, triggers and SLOs, need a separate write permission. Your team must have Honeycomb Intelligence enabled. Its pricing page says the hosted server “is available to all Honeycomb users, including Free.”
  • Elastic. Not a separate service but an endpoint on your own Kibana, {KIBANA_URL}/api/agent_builder/mcp. The docs describe API key authentication, and OAuth 2.1 on Serverless. Tools run with the scope assigned to the key. The older Elasticsearch MCP server is deprecated in favor of this endpoint.
  • Splunk. The Splunk MCP Server app installs on a search head and covers search, alerts and Splunk Observability tools. The 2.0 docs say administrators map roles to tools, and the hosted MCP Gateway is “only available for Splunk Cloud customers in select regions.” Version 2.0.0 was released September 2, 2026.
  • Atlassian. Hosted at https://mcp.atlassian.com/v2/mcp, built for Atlassian products such as Jira and Confluence. Itssupported tools page lists three Jira Service Management tools for operations alerts: getJsmOpsAlerts, getJsmOpsScheduleInfo and updateJsmOpsAlert, which can acknowledge, close or escalate an alert. Those tools work only with an API token that an organization admin has enabled.

Vendors without an official server

A server is counted here only if the vendor publishes it. On October 17, 2026 we searched each vendor’s site, docs and GitHub organization. Community-built servers are not counted and we did not search for them.

  • StatusCake. No MCP repository in its GitHub organization, and a search of statuscake.com for MCP finds nothing.
  • Pingdom. No MCP repository in the SolarWinds GitHub organization and we found no mention on pingdom.com.
  • Uptime Kuma. No official server in the repository. A proposal for one (issue 7795) was closed on August 31, 2026, and a related request (issue 7800) was open on September 2.
  • Instatus. No mention of MCP in its API help pages or its sitemap, and no GitHub repository.
  • Healthchecks.io. No MCP repository in its GitHub organization and no mention on its site, docs or API pages. Its REST API supports an optional read-only key.

Limits and permissions to check before you connect

An MCP server acts with your credentials, so the settings below matter more than the tool count.

  • Pick a read-only option first. UptimeRobot, OpenStatus and Grafana Cloud offer it at sign-in, and Site Qwality does too. Hyperping, HetrixTools and UptimeRobot also support read-only API keys, and Hyperping refuses every write made with one. Rootly’s server has a setting that turns write tools off. Pulsetic says its API key has full access, and HetrixTools says a newly generated key has full access until you scope it.
  • Look for the audit trail. Site Qwality records the assistant and tool name for every change it makes (fields via and mcp_tool), and OpenStatus writes mutations to its audit log with actor_type = "mcp". Hyperping records each write made with an API key in the project audit log, which is on its Business plan. Datadog says all MCP tool calls are recorded in its Audit Trail with the tool name, arguments and user.
  • Know the usage limits. Datadog states fair-use limits of 50 requests per 10 seconds and 100,000 tool calls a month, and New Relic allows 2,000 tool calls per hour per user. UptimeRobot shares the API rate limit. Checkly says trigger-checks consumes check-run quota. Site Qwality applies its normal API limits and quotas.
  • Check where it runs. Datadog does not support GovCloud, Checkly’s public server is not available to accounts that require mTLS, and Grafana’s Cloud server does not work with self-hosted Grafana.

Where another vendor is the better choice

Site Qwality’s server fits teams that want monitors, incidents and status pages in one assistant-ready account without paying extra for it. It is the weaker option in these cases.

  • You want the widest tool surface. Datadog (383 tools listed), Rootly (200+), Better Stack (100+ operations listed) and Cronitor (71) expose more than our 64. Datadog, Grafana, Better Stack and Sentry each document more telemetry query tools than our 5 read-only ones.
  • You need port, ping or TCP checks from the assistant. UptimeRobot, Checkly, Pulsetic and Uptime.com can create them. Site Qwality has no TCP, ICMP or port checks, and its curated tools create only HTTP monitors. Other check types need the generic call_api tool.
  • You want to run the server yourself. Cronitor, Sentry, Grafana, Uptime.com and Rootly publish local servers. Site Qwality’s is hosted only.
  • You need deep on-call and escalation control. PagerDuty’s server creates and updates schedules (including overrides), teams, services and incidents, and reads escalation policies. Grafana’s covers OnCall, incident.io’s can create escalations and change who is on call, and Rootly’s reaches on-call shifts and handoffs. Ours has incident tools and reaches on-call schedules only through call_api.
  • Your security review needs a certification or a region. Site Qwality has no SOC 2 report or ISO 27001 certificate of its own, offers no EU-hosted deployment and runs on one AWS region (us-east-1). Browser, TLS, domain and DNS-record checks run from Virginia only.

FAQ

Yes. As of October 17, 2026, Better Stack, UptimeRobot, Checkly, Cronitor, OpenStatus, Pulsetic, HetrixTools, Hyperping, Uptime.com, Datadog, Sentry, Grafana, PagerDuty, incident.io, Rootly and Site Qwality each publish one, and each documents setup for Claude or Claude Code. Datadog, UptimeRobot, Checkly, Cronitor, OpenStatus, Pulsetic, HetrixTools, incident.io and Site Qwality also document ChatGPT. Which ChatGPT plans can add a custom MCP connector is OpenAI’s rule, and we could not confirm it today. New Relic, Dynatrace, Honeycomb, Elastic, Splunk and Atlassian publish servers too.

All sixteen in the table have write tools, so you restrict them. UptimeRobot, OpenStatus, Hyperping, HetrixTools, Grafana and Site Qwality offer a read-only choice at sign-in or on the key, and Rootly has a setting that turns write tools off. Sentry splits access into skills, of which inspect is read only. Datadog hides write tools with an org setting. Checkly, Cronitor and PagerDuty follow roles, permissions or a scoped OAuth client, and incident.io API keys take chosen scopes. Better Stack uses tool allowlist headers. Pulsetic and Uptime.com document no read-only mode.

Site Qwality includes its server on every plan, including free. HetrixTools says MCP access is on all Uptime Monitor and Blacklist Monitor plans, including Free, and Checkly’s pricing page ticks MCP on all four plans. Honeycomb says its hosted server is available to all users including Free, and New Relic says MCP is available to all accounts. Grafana Cloud’s server is part of Grafana Assistant, whose Free plan includes 40 million tokens per active AI user each month. Pulsetic’s pricing page does not tick it for Free, Hyperping does not include it on Free, and incident.io does not include it on Basic (free). Most other vendors’ MCP docs do not state plans.

It depends on the vendor. UptimeRobot says delete operations are not exposed over MCP, and Hyperping says agents never delete monitors, status pages or incidents, with one exception: canceling a maintenance window that has not started. Pulsetic has a delete_monitor tool, Checkly has delete-check, Uptime.com has delete_check, and Site Qwality has delete_monitor, which it flags as destructive. Where the vendor offers a read-only option, you can withhold write access.

Cronitor (inside CronitorCLI), Sentry (@sentry/mcp-server), Grafana (mcp-grafana), Uptime.com (a Go program or container) and Rootly (uvx or Docker) publish local servers. PagerDuty’s and Dynatrace’s local servers are deprecated in favor of their hosted ones. Site Qwality’s server is hosted only.

Not an official one that we could find. As of October 17, 2026 we found no MCP repository in their GitHub organizations and no mention on their sites or docs. Uptime Kuma has a closed proposal and an open request on GitHub. We did not search community-built servers.

It lets an assistant list and pause monitors, read check results, work incidents, post status page updates and search logs, metrics and traces, as you and within the access you approve. It is hosted at https://mcp.siteqwality.com/mcp, signs in with OAuth or an API key, and records every change in your audit log. See the MCP server page for the full tool list.

Fonti

Ogni dato sui concorrenti in questa pagina proviene dalle pagine ufficiali del fornitore.

Altri confronti

Site Qwality vs Better Stack Prezzo, funzionalità e profondità del monitoring a confronto. Site Qwality vs Datadog Prezzo, funzionalità e profondità del monitoring a confronto. Site Qwality vs UptimeRobot Prezzo, funzionalità e profondità del monitoring a confronto. All-in-one monitoring Prezzo, funzionalità e profondità del monitoring a confronto.

Inizia gratis. Nessuna carta di credito.

Uptime, cron, synthetic, log, RUM, incidenti e pagine di stato. Piano gratuito su ogni prodotto.

Inizia gratis Vedi i prezzi