Back to home
MCP Server

Your monitoring, inside your AI tools

Vantaj ships a hosted Model Context Protocol server. Ask Claude which monitors are down, have Cursor create a monitor for the endpoint you just built, or let an agent pull incident history while debugging - authenticated with your existing API key.

Hosted, nothing to run

A remote MCP endpoint at api.vantaj.co/mcp over Streamable HTTP. No local server, no npm package, no OAuth dance - paste a URL and a header.

Works with Claude, Cursor, and more

Any MCP-capable client connects: Claude, Claude Code, Cursor, ChatGPT, or your own agents. Local-only clients bridge via mcp-remote.

Your API keys, your scopes

Authentication reuses your existing Vantaj API keys. Scoped permissions apply exactly as they do on the REST API - a read-only key gives your AI read-only tools.

Safe by design

Reads and additive writes only. No delete tools, no alert-policy or billing access - a manipulated model can't destroy anything.

Full monitoring context

Monitors, heartbeats, incidents, domains, and status pages - your assistant sees live status, uptime, and history across all of it.

Create monitors in chat

Ship an endpoint, then tell your assistant to monitor it - including inverted exposure checks. Checks start immediately, plan limits enforced.

Available tools

ToolWhat it does
list_projects
Discover projects - the entry point most other tools build on
list_monitors / get_monitor
Live status, response times, and uptime for your monitors
create_monitor / update_monitor
Create and edit monitors, including inverted exposure checks
pause_monitor / resume_monitor
Stop and restart checks for a monitor
create_maintenance_window
Suppress alerts during planned work - let your deploy pipeline silence monitors before shipping
list_heartbeats
Cron and background-job monitors with last-ping status
list_incidents / get_incident
Open and resolved incidents with timestamps and duration
list_domains / list_status_pages
Domain expiry countdowns and status page configuration

How it works

01 - Create an API key

Same keys as the REST API

Generate a key under Settings → API Keys and give it the scopes you want your assistant to have. Read-only keys make the whole MCP surface read-only.

02 - Add the connector

One URL, one header

Point your client at api.vantaj.co/mcp with an Authorization header. Claude, Claude Code, and Cursor each take about 30 seconds to configure.

03 - Ask away

Your monitoring joins the conversation

Ask "which monitors are down?", "what incidents did we have this week?", or "monitor https://api.acme.com/health every minute" - the assistant calls the right tools.

04 - Stay in control

Guardrails built in

Every call runs through the same scopes, plan limits, and rate limits as the REST API - and destructive operations simply don't exist on this surface.

Frequently asked questions

Which AI clients are supported?
Anything that speaks MCP over Streamable HTTP: Claude (web and desktop), Claude Code, Cursor, ChatGPT connectors, and custom agents. Clients that only support local stdio servers can bridge with the mcp-remote package.
Can an AI delete my monitors?
No. The MCP surface deliberately has no delete tools and no access to alert policies, team, or billing settings. Destructive actions stay in the dashboard and REST API, where a human is in the loop.
Does MCP access cost extra?
No. It's included with API access, which is available on paid plans. MCP calls count toward the same rate limits as regular API requests.
How is it authenticated?
With your regular Vantaj API key sent as an Authorization: Bearer header. Keys are scoped, so you control exactly what a connected assistant can read or change.
Is there a local server to install?
No - the server is hosted at api.vantaj.co/mcp. There's nothing to run, update, or patch on your side.

Ready to set up MCP Server?

Up and running in under a minute. Free for up to 20 monitors, no credit card required.

Start for free