The Hyperping MCP server gives Claude, Cursor, Codex and any MCP client live access to your monitors, outages, uptime, on-call and status pages. Ask in plain words, or let it create monitors, page on-call and post incidents for you.
On Essentials, Pro and Business. Nothing to install.
What’s down right now?
get_status_summary()get_outage_timeline(uuid: "outage_8f2c")2 of your 48 monitors are down.
2 down · 46 up
Checkout went down at 14:02 UTC, confirmed from London, Frankfurt and Virginia. Sarah acknowledged it two minutes later. Want the full timeline?
“Why is checkout down?” brings back the timeline: detection, confirmation from other regions, alerts sent and who acknowledged. Follow-ups drill into the probe logs.
A problem no monitor catches? Your agent declares the incident and pages the on-call responders of the escalation policy you pick, then acknowledges, escalates or resolves it when you ask.
Your agent drafts the incident from what your monitors saw and publishes it on your status page once you approve the wording. Updates and the resolution follow the same way, and subscribers get each one.
Your agent creates the page on a hyperping.app address with the services you name, then changes its text, theme, colors or subscription button when you ask. Custom domains, logos and access control stay in the dashboard.
Schedule the window and announce it on your status page, with a heads-up to subscribers. Checks and alerts stop for the monitors involved, then start again on their own. If plans change, your agent reschedules it, ends it early or cancels it.
Uptime, response times, MTTR and MTTA per monitor over any window, ready for a weekly digest or a customer’s SLA review.
Each scheduled run pings /start, then pings again only when its output passes your check. A rate limit, a loop cut by its timeout or an empty answer sends nothing, and Hyperping alerts you.
Every healthcheck with its state, its last ping, when the next one is due and a late flag during its grace period. Enough for a morning brief of what’s down, new or late.
The last 20 pings with each run’s duration, 30 days of uptime and the recent incidents. An agent can also check that its upstream job ran before it starts.
list_monitorsFilter by up, down, paused or SSL expiring.get_monitorOne monitor with its full configuration.search_monitors_by_nameFind monitors by name or URL.create_monitorwriteHTTP, ping, port or DNS checks.update_monitorwriteChange any setting of a monitor.pause_monitorwriteStop checks and alerts, keep history.resume_monitorwriteStart checking again.get_status_summaryWhat’s up, what’s down, since when.get_monitor_uptimeSLA over any window, per day to month.get_monitor_response_timeLatency trend per monitor.get_monitor_mttrMean time to resolve.get_monitor_mttaMean time to acknowledge.list_outagesDetected or declared, searchable and filtered.get_outageOne outage and its details.get_outage_timelineDetection to resolution, step by step.get_monitor_outagesThe outages of one monitor.create_outagewriteDeclare one by hand, page on-call.acknowledge_outagewriteMark it handled, stop repeat alerts.escalate_outagewritePage the next escalation step now.resolve_outagewriteResolve it and send the recovery.get_monitor_anomaliesFlapping and latency spikes.get_monitor_http_logsRecent probe logs.list_recent_alertsAlerts sent, across every channel.list_on_call_schedulesEvery rotation.get_on_call_scheduleOne rotation and its overrides.list_escalation_policiesEvery escalation chain.get_escalation_policyOne policy, step by step.list_team_membersNames and emails of your team.list_integrationsSlack, Teams, PagerDuty, webhooks…get_integrationOne channel and its settings.list_status_pagesEvery page, its URL and its access.get_status_pageIts languages, subscriptions and services.create_status_pagewriteA public page on a hyperping.app address.update_status_pagewriteName, text, look and subscription button.add_status_page_serviceswriteShow monitors and components on it.remove_status_page_serviceswriteTake services off the page.list_status_page_incidentsOngoing and resolved, newest first.get_status_page_incidentOne incident and every update.create_status_page_incidentwritePublish it and notify subscribers.add_status_page_incident_updatewriteIdentified, monitoring or a new update.resolve_status_page_incidentwriteClose it with a final update.update_status_page_incidentwriteTitle, type, pages or affected services.edit_status_page_incident_updatewriteFix a posted update without notifying again.list_maintenance_windowsUpcoming, ongoing and past windows.get_maintenance_windowOne window and its announcement.create_maintenance_windowwriteSchedule it, announce it, pause checks.update_maintenance_windowwriteReschedule it, change it, post an update.complete_maintenance_windowwriteEnd it now, checks resume.cancel_maintenance_windowwriteDelete one that hasn’t started.In the Hyperping sidebar, open API Keys. Pick read-only for agents that only need to look, read and write for the ones that act.
One URL and one header. Pick your client for the exact command or file, or paste the JSON into any MCP client.
“What’s down right now?” is a good first question. Your agent picks the tools, Hyperping returns the facts.
claude mcp add --transport http hyperping \
https://api.hyperping.io/v1/mcp \
--header "Authorization: Bearer $HYPERPING_API_KEY"Each API key belongs to a single project. Anything outside it returns a 404, even with a valid ID.
Give agents you don’t fully trust a read-only key. Write tools refuse it in the server, so no prompt can talk its way around it.
Each tool carries MCP annotations that say whether it only reads, can overwrite or remove data, or reaches people outside the dashboard. Clients such as Claude Code read them to decide which calls run on their own and which ask you first.
Agents can create and change monitors, status pages, incidents and maintenance windows, not delete them. The only exception is canceling a maintenance window that hasn’t started, which deletes it. Everything else is deleted in the dashboard or the REST API.
On the Business plan, the project’s Audit Log lists each change an agent makes with an API key, under that key, as it does for REST API writes. A read-only key that tries to write is logged too.
Delete a key in the dashboard and the next call fails. Keys can also expire on their own, from one hour to one year.
USD · Monthly billing for every plan
Three subscriptions to keep it all running.
$747/month
$8,964 over 12 months
Everything connected, from the first check.
$299/month
$3,588 over 12 monthsSave $5,376 a year with this mix.
Reviewed . All amounts are USD; local currency prices may differ.
This is one example of a three-tool stack, not the cheapest possible setup or a feature-for-feature match. Pingdom and PagerDuty also include status-page features; a separate Statuspage subscription may not be needed by every team. Usage charges and paid connectors are not included. Annual totals are 12 monthly payments, not annual subscription quotes.
A hosted Model Context Protocol server at https://api.hyperping.io/v1/mcp. It exposes a Hyperping project to AI agents as 49 tools: 28 read tools for monitors, incidents, alerts, uptime, response times, MTTR, MTTA, on-call, integrations, status pages, status page incidents and maintenance windows, and 21 write tools to manage monitors, declare, acknowledge, escalate and resolve incidents, create and edit status pages and their incidents, and schedule, reschedule, end or cancel maintenance. See the tool reference.
Any MCP client that connects to remote servers over streamable HTTP with a custom header: Claude Code, Cursor, OpenAI Codex, VS Code with GitHub Copilot, Windsurf, Gemini CLI and Zed. Clients that only run local servers, such as Claude Desktop, connect through the mcp-remote bridge. The agent setup guide has the configuration for each.
The MCP server uses API keys, available on the Essentials, Pro and Business plans. MCP calls cost nothing extra. See pricing.
With a Hyperping API key sent as an Authorization: Bearer header. Each key belongs to one project and is either read-only or read and write. OAuth sign-in is not supported yet: clients that only offer OAuth connect through the mcp-remote bridge.
No. Deletion is not exposed over MCP, with one exception: canceling a maintenance window that hasn’t started deletes it. An agent with a read and write key can create and change monitors, status pages, incidents and maintenance windows, but deleting a monitor, a status page, an incident or an update takes the dashboard or the REST API. A read-only key cannot change anything.
Yes, with a read and write key. It can create a status page, change its settings and the services it shows, publish an incident, post and edit its updates, resolve it, and schedule, reschedule, complete or cancel maintenance. New incidents and updates reach subscribers as when you post from the dashboard, unless the agent turns notifications off, and a maintenance window notifies them only if the agent asks. Editing an incident, an update or a maintenance window notifies no one. Password protection, SSO, custom domains and logos stay in the dashboard. The server asks agents to show you the exact text before publishing, and a read-only key cannot publish anything.
Yes, with a read and write key. It can declare an incident for a problem no monitor detects and page the on-call responders of an escalation policy, acknowledge an incident so repeat alerts stop, page the next escalation step right away, and resolve a declared incident. These incidents stay internal: nothing appears on a status page unless the agent also publishes a status page incident. Before paging anyone, the server asks agents to show you the text and who will be paged.
Yes, with a read-only key. list_healthchecks returns each healthcheck’s state, last ping, the time the next ping is due and a late flag while it is in its grace period, and get_healthcheck adds its last 20 pings and recent incidents. The guide to monitoring scheduled AI agents builds a morning brief on them.
Yes, on the Business plan. Every change an agent makes with an API key appears in the project’s Audit Log under the key that made it, as for REST API writes, and a read-only key that tries to write is logged too. Each tool also carries MCP annotations, which clients such as Claude Code read to decide which calls run on their own and which ask you first.
Per project, 60 tool calls a minute and 600 an hour. Calls also count toward the project’s REST API quota. Past a limit, the server returns an error that says how many seconds to wait.
Yes. /docs/llms.txt indexes every page, every page has a Markdown version at its URL followed by index.md, and /mcp/llms.txt describes every MCP tool with its parameters and common queries.
