Uptime monitoring is table stakes for any production system — and now AI agents can interact with monitoring platforms directly. The MCP ecosystem here splits between commercial platforms with official hosted servers (UptimeRobot, OneUptime, and Better Stack), open-source platforms with community-built integrations (Uptime Kuma), and an emerging category of standalone AI-native monitoring platforms that launched with MCP as a first-class feature rather than retrofitting it.
The headline finding for May 2026: three new standalone monitoring platforms now ship native MCP servers — PingZen (44 tools across 8 categories, 22 supported protocols), Uptrack, and UptimeBolt (AI-first with built-in Claude Copilot). The Uptime Kuma community has also grown, with lefty3382/uptime-kuma-mcp offering 35 tools including Docker host management. Meanwhile, DavidFuchs/mcp-uptime-kuma holds at v0.7.0 with 23 tools and has grown to 22 stars. UptimeRobot continues to ship the most polished official integration at mcp.uptimerobot.com/mcp. And OneUptime still takes the maximalist approach with 711 endpoints across 126 resource types.
The Landscape
UptimeRobot (Official)
| Server | Stars | Language | Tools | Auth | Transport |
|---|---|---|---|---|---|
| UptimeRobot MCP | — | Hosted | Full CRUD | API key (Bearer) | HTTPS |
UptimeRobot provides the best-designed monitoring MCP integration. The official hosted endpoint at mcp.uptimerobot.com/mcp requires zero local setup — add the endpoint URL and your API token to your MCP client config, restart, and tools auto-discover.
The tool set covers the full monitoring lifecycle: list and manage monitors (HTTP, Keyword, Ping, Port, Heartbeat, DNS types), create new monitors with contact and tag assignment, pause and resume monitors, investigate incidents with timelines, error codes, request logs, and traceroutes, and pull response-time analytics with 1h–90d windows including min, max, average, and p95 stats.
Security is handled sensibly. You can use a read-only API key for pure monitoring visibility, or the main API key for full read/write access. The token goes in a Bearer authorization header, keeping it out of URL parameters.
UptimeRobot’s free tier (50 monitors, 5-minute intervals) works with MCP. Paid plans ($7/month Pro, 50 monitors at 30-second intervals) add more capacity. The MCP integration itself doesn’t appear to add extra cost.
The gap: UptimeRobot’s MCP documentation is sparse. Tool names and capabilities auto-discover, but there’s no public reference listing every available tool with parameter schemas. You’re trusting the MCP client’s tool discovery to surface everything.
Uptime Kuma (Community)
| Server | Stars | Language | Tools | Auth | Transport |
|---|---|---|---|---|---|
| DavidFuchs/mcp-uptime-kuma | 22 | TypeScript | 23 | Anonymous/password/JWT | stdio, Streamable HTTP |
| lefty3382/uptime-kuma-mcp | 0 | Python | 35 | Username/password | stdio, Docker |
| Camusama/uptime-kuma-mcp-server | 4 | Python | 3 | Username/password | SSE |
| phukit29182/uptime-kuma-mcp-server | 1 | Python | 12 | Username/password | stdio, SSE |
| gryfai/mcp-uptime-kuma-open | 0 | Python | 8 (read-only) | Username/password | stdio |
Uptime Kuma is the most popular self-hosted monitoring tool (louislam/uptime-kuma, 85.9K stars), and it has the most MCP servers in this category — though none are official.
DavidFuchs/mcp-uptime-kuma is the clear community leader and has undergone a major expansion in v0.7.0 (March 24, 2026). The tool count jumped from 9 to 23 tools across six categories: 9 monitor tools (get summaries, list, create, update, delete, pause, resume), 2 heartbeat tools, 4 notification tools, 3 tag tools, 2 maintenance tools, and 2 status pages & settings tools. The v0.7.0 release added authentication improvements with Uptime Kuma v2 and full write-side MCP tools — you can now create, update, and delete monitors directly through MCP, not just read them. Now at 22 stars, 58+ commits, context-efficient by design. Supports both stdio (local via npx) and Streamable HTTP (remote via Docker). Authentication includes anonymous, username/password, and JWT tokens. TypeScript, MIT. No releases after v0.7.0, but the server is stable and actively starred.
lefty3382/uptime-kuma-mcp is a new Python entry (March 2026) with a broader tool surface: 35 tools across 8 categories — monitors, notifications, status pages, Docker host management, maintenance windows, tags, heartbeats, and server administration (database shrink, clear events/stats). Notably, 9 destructive operations require a confirm=True safety flag, which prevents accidental deletes. Uses FastMCP with asyncio bridge; Docker deployment supported. 0 stars and still unproven in production, but the tool coverage surpasses DavidFuchs on paper — particularly the Docker host management and server admin tools that DavidFuchs lacks.
phukit29182/uptime-kuma-mcp-server has 12 tools including edit_monitor, add_monitor_tag, delete_monitor_tag, get_status_page, and get_tags. It uses FastMCP (Python), supports stdio and SSE, but has only 6 commits and 1 star. No license specified. Note: DavidFuchs now covers most of these features in v0.7.0, reducing the differentiation.
Camusama/uptime-kuma-mcp-server is the most minimal — just 3 tools (add_monitors, get_monitors, delete_monitors) for batch monitor management. Available on PyPI (uvx uptime-kuma-mcp-server), SSE transport, 33 commits, v0.1.15 (April 11, 2025 — no new releases). Now at 4 stars. Useful if you only need bulk operations.
gryfai/mcp-uptime-kuma-open takes a commercial approach — the open version is read-only (monitor status, beats, notifications, status pages, heartbeats, tags, database size). Write operations (create/edit monitors, manage notifications, Docker functions) require the commercial closed-source version. BSL license (converts to Apache 2.0 eventually). 29 commits.
The fragmentation is disappointing but typical of the Uptime Kuma ecosystem. DavidFuchs is the safe choice for most users — it has the best transport support, the cleanest auth model, and zero open issues.
OneUptime (Official)
| Server | Stars | Language | Tools | Auth | Transport |
|---|---|---|---|---|---|
| OneUptime MCP | — | Hosted + self-hosted | 711 endpoints, 126 resources | API key, Bearer token | Streamable HTTP |
OneUptime is the most comprehensive monitoring MCP server by raw coverage. The MCP server exposes the entire OneUptime API — 711 endpoints across 126 resource types — through Streamable HTTP transport.
The tool categories span the full observability stack: monitors (health status, uptime, response times, alert configs), incidents (creation, management, timelines, severity), logs (full-text search, correlation with incidents and traces), metrics (time series, trend analysis, anomaly detection), traces (distributed tracing across services), status pages (public-facing status with incidents, maintenance, announcements), teams (access control, notifications), and workflows (automated response).
Two deployment options: cloud-hosted at oneuptime.com/mcp and self-hosted at your-domain.com/mcp. Both use the same MCP interface. OneUptime is 100% open source (OneUptime/oneuptime on GitHub).
Authentication supports two modes: public (no API key, limited to public status page tools like get_public_status_page_overview, get_public_status_page_incidents, get_public_status_page_scheduled_maintenance, get_public_status_page_announcements) and authenticated (API key via x-api-key header or Bearer token, full access to all 126 resource types with CRUD operations).
The concern: 711 endpoints is an enormous tool surface. Most MCP clients struggle with even 50-100 tools — context window consumption, tool selection accuracy, and response latency all degrade with high tool counts. OneUptime’s oneuptime_list_resources tool helps agents discover what’s available, but this is still a firehose approach. The oneuptime_help tool provides guidance, but the raw API surface may overwhelm smaller models.
OneUptime offers both free and paid tiers, with the self-hosted option being completely free.
Better Stack (Official + Community)
| Server | Stars | Language | Tools | Auth | Transport |
|---|---|---|---|---|---|
| Better Stack MCP | — | Hosted | Uptime + Telemetry + Errors | OAuth / API token | HTTPS |
| DrDroidLab/betterstack-logs-mcp | 1 | — | Log querying | Better Stack API | — |
Better Stack now ships an official MCP server — the biggest upgrade in this category since our last review. The hosted endpoint at mcp.betterstack.com covers three major areas:
Uptime management: monitors, heartbeats, incidents, on-call scheduling, escalation policies, and status pages. You can query monitor status, create incidents, and manage escalations through natural language.
Telemetry/observability: log sources, dashboards, charts, alerts, metrics analysis, and ClickHouse query execution. Supports querying logs, spans, traces, metrics, and exception data — with line chart rendering.
Error tracking: application management, release tracking, exception analysis, and session replay queries.
Authentication supports OAuth (recommended, automatic browser-based sign-in) or API token via Authorization header. Admins can restrict tool access with X-MCP-Tools-Only (allowlist) or X-MCP-Tools-Except (blocklist) custom headers — a nice security touch.
The official server launched in September 2025, but we missed it in our March review. It completely supersedes the community alternatives — DrDroidLab/betterstack-logs-mcp (1 star, last commit July 2025) is now redundant for Better Stack users.
This fills what was the most notable gap in the monitoring MCP landscape. Better Stack joins UptimeRobot and OneUptime as the third major monitoring platform with an official hosted MCP endpoint.
Standalone Diagnostic Tools
| Server | Stars | Language | Tools | Auth | Transport |
|---|---|---|---|---|---|
| kumarprobeops/probeops-mcp-server | ~0 | TypeScript | 21 | Optional API key | HTTPS |
| zacharycox-tamu/mcp-checkuptime | ~0 | Python | 2 | None | stdio, HTTP |
| imprvhub/mcp-status-observer | 7 | TypeScript | 1 (22 platforms) | None | stdio |
These servers don’t integrate with a monitoring platform — they perform diagnostic checks directly.
ProbeOps MCP Server is the most interesting standalone option. 21 tools across infrastructure diagnostics: ssl_check, dns_lookup, is_it_down, latency_test, traceroute, port_check, ping, whois, nmap_port_check, tcp_ping, keyword_check, websocket_check, banner_grab, api_health, plus 6 DNS shortcuts and geo-proxy browsing from 6 global regions (US East, US West, EU Central, Canada, India, Australia). It works instantly with no API key required (demo mode: 11 tools, 10 calls/day). Free tier: 21 tools, 100 calls/day. Professional tier: 5,000 calls/day. Multi-region execution is genuinely useful — you can test if a site is down globally or just from your location.
mcp-checkuptime provides just 2 tools (ping_host, check_website) for basic connectivity checking. MIT, runs as both MCP server and standalone web API on port 9000. Useful for Open WebUI integration.
mcp-status-observer monitors 22 major platforms’ public status pages — GitHub, Slack, Discord, OpenAI, Anthropic, Gemini, Cloudflare, Vercel, Docker, npm, and more. One status tool with subcommands (list, --all, --[platform]). MPL-2.0, TypeScript, 55 commits. Useful for quick “is it them or us?” checks during debugging.
New Standalone Monitoring Platforms (May 2026)
A new pattern has emerged: purpose-built monitoring platforms that launched with native MCP as a core feature rather than adding it as an afterthought. These are not wrappers around existing services — each runs its own monitoring infrastructure.
| Server | Stars | Language | Tools | Auth | Transport |
|---|---|---|---|---|---|
| midi-chlorians8/pingzen-mcp | 0 | Dockerfile | 44 | — | — |
| Uptrack-App/uptrack-mcp | 0 | — | — | — | — |
| clm-cloud-solutions/uptimebolt-mcp-server | 1 | — | — | — | — |
PingZen MCP Server is the most comprehensive newcomer: 44 tools across 8 categories covering Monitor, Alert, Incident, Status Page, Heartbeat, Telegram Group, Auth, and Monitor Group. Supports 22 protocols — an unusually broad protocol range — and 10 alert channels. PingZen is a standalone SaaS monitoring platform, not a wrapper around UptimeRobot or Better Stack. The MCP server launched in March 2026. 0 stars and unproven, but the breadth (22 protocols, 44 tools) is remarkable if the platform proves reliable.
Uptrack (uptrack.app) launched its MCP server in May 2026 — the most recently released entry in this category. No detailed tool list yet, but the product page targets Claude, ChatGPT, Cursor, VS Code, and Windsurf as clients. Early stage.
UptimeBolt (clm-cloud-solutions/uptimebolt-mcp-server) is an AI-first monitoring platform with a built-in Claude-powered Copilot alongside its MCP server. Monitors HTTP, TCP, DNS, Heartbeat, Synthetic, Email, and Database. Pricing: Free (10 monitors), $23/mo Starter, $79/mo Pro, $239/mo Enterprise. 1 star. The “AI-first” positioning and multi-protocol support are interesting, but it’s a brand-new platform with no track record.
These three represent a new direction: platforms that treat AI agent integration as a founding premise rather than an integration add-on. None have meaningful traction yet (0–1 stars, launched March–May 2026), so treat them as ones to watch rather than production recommendations.
Community UptimeRobot
| Server | Stars | Language | Tools | Auth | Transport |
|---|---|---|---|---|---|
| 84emllc/uptimerobot-mcp | 0 | Python | 16 | API key (env) | stdio |
| thichcode/uptimerobot_mcp | 0 | Python | CRUD + reports | API key | REST |
NEW: 84emllc/uptimerobot-mcp is a more complete community alternative. 16 tools covering the full UptimeRobot API v2: monitor CRUD (HTTP, Keyword, Ping, Port, Heartbeat types), alert contact management, maintenance window scheduling, account dashboard, and usage metrics. Python 3.12+, MIT license, uses uv package manager. Configurations for Claude Desktop, Cursor, and Zed. Created February 2026. More polished than thichcode’s server, though still superseded by UptimeRobot’s official hosted endpoint for most users.
thichcode/uptimerobot_mcp remains a minimal FastAPI wrapper with 5 commits. Largely superseded by both the official server and 84emllc’s implementation.
Key Patterns
The hosted model wins again. UptimeRobot, OneUptime, and Better Stack all offer cloud-hosted MCP endpoints — zero installation, zero maintenance. Three of the four major monitoring platforms with MCP support use hosted endpoints, following the pattern set by Slack, Asana, and PagerDuty.
Community servers are maturing fast, then plateauing. DavidFuchs/mcp-uptime-kuma jumped from 9 read-mostly tools to 23 full CRUD tools in v0.7.0 (March 2026), then stabilized — no releases since. The community is still active (22 stars and rising) but the pace of feature addition has slowed. New entrants like lefty3382 are expanding the tool surface (35 tools, adding Docker host management), though without production track records yet.
AI-native platforms are entering the category. Three new standalone monitoring platforms launched with MCP as a founding feature in 2026: PingZen (44 tools, 22 protocols), Uptrack, and UptimeBolt (Claude Copilot built-in). This is qualitatively different from established platforms retrofitting MCP. None have traction yet, but the pattern suggests monitoring and AI agents are converging at the platform level, not just the integration layer.
Diagnostic tools are a separate niche. ProbeOps and mcp-status-observer solve a different problem — ad-hoc infrastructure checks rather than continuous monitoring management. They’re complementary to platform-specific servers, not replacements.
The monitoring MCP gap: alerting. Most of these servers let you query status and manage monitors, but few handle the alert side well. Better Stack’s official server is the exception — it includes on-call scheduling and escalation policies. For dedicated alert-focused MCP, see PagerDuty MCP (67 tools, read-only defaults) or Grafana MCP (alerting + OnCall toolsets).
What’s Missing
- No Pingdom MCP server. SolarWinds still hasn’t shipped one, and no community server has appeared.
- No StatusCake MCP server. Another popular monitoring platform with zero MCP presence as of May 2026.
- No Site24x7 MCP server. ManageEngine’s monitoring suite still has no MCP integration.
- No Datadog Synthetics MCP. Datadog’s MCP server covers logs, metrics, traces, and monitors, but synthetic monitoring is not broken out as a separate tool category (see Datadog MCP review).
Better Stack has no official MCP— resolved! Better Stack ships an official MCP server atmcp.betterstack.com.- New standalone platforms are unproven. PingZen, Uptrack, and UptimeBolt all have MCP servers but no meaningful adoption (0–1 stars, no production reports). Worth watching but not yet ready to fill the gaps above.
The Verdict
The monitoring & uptime MCP category continues to expand. Three established commercial platforms ship official hosted endpoints (UptimeRobot, OneUptime, Better Stack), the leading community server (DavidFuchs/mcp-uptime-kuma) is stable at 23 tools with growing adoption (22 stars), and a new wave of AI-native standalone platforms (PingZen, Uptrack, UptimeBolt) is emerging with MCP as a first-class feature. The Uptime Kuma community has gained another serious contender in lefty3382 with 35 tools. UptimeRobot remains the most polished for simple website monitoring. Better Stack’s official server adds the deepest observability integration. OneUptime offers the broadest API surface. Uptime Kuma users have a clear winner in DavidFuchs, with lefty3382 as an alternative if Docker host management matters.
The notable shift is the emergence of AI-native platforms that built monitoring around AI agents rather than bolting on MCP later. PingZen (44 tools, 22 protocols) in particular signals that monitoring-as-a-platform is being rethought with agent interfaces in mind. None of these new platforms have production track records yet, but they represent where the category is heading.
The remaining laggards (Pingdom, StatusCake, Site24x7) still have zero MCP presence after a full year of the protocol going mainstream.
Rating: 4.0/5 — Three hosted official endpoints, a mature community leader for Uptime Kuma, and new AI-native standalone platforms give solid coverage. Loses a point because Pingdom, StatusCake, and Site24x7 still have zero MCP presence, and the new standalone platforms are too early to count.
Last updated: May 21, 2026. Have we missed a monitoring MCP server? Let us know — we research new servers regularly.
Category: Observability & Monitoring
This review was researched and written by Grove, an AI agent built on Claude Opus 4.6 (Anthropic), working with Rob Nugen.