Bring Hyperping into your AI workflows.

Connect Claude, Cursor, Windsurf, or any MCP-compatible client to your Hyperping project with one API token. Read monitors, drill into outages, follow the on-call chain, and pause services from a prompt.

Your whole Hyperping project, exposed as tools.

Your AI composes the queries; Hyperping returns the facts. Deletion is intentionally not exposed. Reversible changes only.

Three steps. One token.

Works with any MCP-compatible client. No local server, no SDK.

1. Create an API key

   In Hyperping, open Project Settings → API Keys. Pick read-only or read-write. Keys are strictly scoped to one project.

2. Drop in the server config

   Paste this into your client's MCP config. The same JSON works for Claude Desktop, Cursor, Windsurf, Cline, and Claude Code.

   ~/.claude/claude_desktop_config.jsonjson

   {
     "mcpServers": {
       "hyperping": {
         "type": "http",
         "url": "https://api.hyperping.io/v1/mcp",
         "headers": {
           "Authorization": "Bearer sk_live_..."
         }
       }
     }
   }

3. Restart and ask

   Restart your client. Start asking questions. Every call reuses the same auth, permissions, rate limits, and activity logs as the REST API.

How the server behaves.

Same guarantees as the REST API. No new abstractions to reason about.

Endpoint

https://api.hyperping.io/v1/mcp

Transport

MCP over HTTP, stateless. No persistent connection.

Auth

Authorization: Bearer sk_live_...

Scope

One API key, one project. No cross-project access.

Write surface

pause, resume, create, update. Deletion is not exposed.

Read-only keys

Supported. All write tools return a 403 for keys without write scope.

Rate limit

300 req / min / key, shared with REST.

Observability

Every call appears in the project's activity log with the tool name, caller, and latency.

What teams actually do with it.

The MCP server turns monitoring from a dashboard you check into a context your AI tools already have.

Incident triage in Claude

"Why is checkout down?" pulls the outage timeline, recent HTTP logs, and response-time history in one conversation — no dashboard tab, no copy-paste. Follow-ups drill in.

Ship & monitor from Cursor

Deploying a new service? Ask your editor's agent to create the monitor, set the check frequency, and wire the escalation policy — while the deploy runs.

Post-mortem drafts

Feed the outage timeline straight into a post-mortem draft: impact window, MTTR, alert dispatches, who acked. Pair it with the post-mortem generator.

Autonomous agents, safely

Give ops agents a read-only key: they can observe, correlate, and report, but every write returns a 403. Full audit trail either way.

Weekly reliability reports

"Summarize this week: uptime, outages, MTTR by monitor, anything anomalous" — a scheduled agent turns it into a Slack digest your team actually reads.

Monitoring as code, reviewed by AI

Combine with the Terraform provider: agents propose monitor changes as HCL, humans merge the PR.

What people actually ask.

Copy any of these into your MCP-connected client.