ObserveOne runs a hosted MCP (Model Context Protocol) server, so an AI agent can read and manage your monitoring directly. Ask Claude, Cursor, opencode, Codex, or ChatGPT "are any monitors down?", "show me open incidents", or "pause the staging check" and the agent answers or acts on your real account.
An agent gets read access plus safe writes. It can list monitors, read uptime, and read incidents, and it can create, update, pause or resume, and run monitors and checks, manage alert channels and status pages, and open, update, or resolve incidents. It cannot delete anything.
What the agent can do#
The agent has read tools across every resource (monitors, API checks, heartbeats, alert channels, status pages, incidents, and Autopilot suites) plus uptime and incident summaries, and safe-write tools to create, update, toggle, and run them. For example:
observeone_monitors_listlists your monitors with type, status, and uptime.observeone_url_monitor_createadds a URL monitor;observeone_api_check_updateedits a check.observeone_incident_createopens an incident;observeone_incident_resolvecloses it.observeone_api_check_runruns a check on demand.
There are no delete tools: an agent can create and change things, but never destroy them.
Connect your client#
The server URL is https://mcp.observeone.com/mcp. Add it in your MCP client's settings. No API key or config snippet is needed.
When you first connect, your client opens a browser window to the ObserveOne authorization page. Sign in to your account and click Authorize Access. The client stores the token and reconnects automatically when it expires.
Claude#
In Claude Desktop or claude.ai, open Settings ▸ Integrations, add a new MCP server, and paste the URL above. On first use Claude opens a browser sign-in. After you approve, the connection is active.
Other clients#
Any MCP client that supports OAuth for remote servers works the same way: add the server URL, approve the sign-in prompt that appears, and you're connected. Check your client's documentation for where to add a remote MCP server URL.
Revoke access#
To see and remove connected agents, open Settings ▸ Integrations. The MCP connections list shows every agent connected over MCP, with its name, scopes, and expiry. Click Revoke on a connection to cut off its access on its next request. The agent has to reconnect before it can use ObserveOne again.
Next steps#
- API Checks: Overview: build the monitors your agent will read.
- Scheduling, alerts & import/export: promote a check to a monitored, multi-region run.