WatchFor

Connect AI agents (Claude, Cursor, VS Code) to WatchFor via the Model Context Protocol — 26 tools over your monitors, incidents and alerting.

WatchFor ships a built-in MCP server, so AI agents can monitor your infrastructure, investigate incidents and manage monitors through natural language. It exposes 26 tools backed by the REST API — the same authentication, permissions and per-plan rate limits apply.

Endpoint: https://watchfor.io/api/mcp (Streamable HTTP)

Authentication uses your API key as a Bearer token. A read-scope key allows only the read tools; write scope enables creating monitors, alert rules, maintenance windows and incident actions.

Connect a client

Claude Code

claude mcp add --transport http watchfor https://watchfor.io/api/mcp \
  --header "Authorization: Bearer wf_live_YOUR_KEY"

Cursor / VS Code / other JSON-config clients

{
  "mcpServers": {
    "watchfor": {
      "url": "https://watchfor.io/api/mcp",
      "headers": {
        "Authorization": "Bearer wf_live_YOUR_KEY"
      }
    }
  }
}

Claude Desktop (via mcp-remote)

Claude Desktop's config takes stdio servers — bridge with mcp-remote:

{
  "mcpServers": {
    "watchfor": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://watchfor.io/api/mcp",
        "--header",
        "Authorization: Bearer wf_live_YOUR_KEY"
      ]
    }
  }
}

Available tools

Tool	Scope	What it does
`get_summary`	read	Org snapshot: monitor status counts, active incidents, health score
`get_incident_stats`	read	Period stats: severity totals, MTTR, MTBF, most unstable monitors
`get_plan`	read	Plan limits and live usage
`list_monitors` / `get_monitor`	read	Browse monitors (filter by type/status/tag)
`get_monitor_types`	read	Discovery catalog: all 25 types, metrics, config fields
`get_monitor_uptime` / `get_monitor_checks`	read	Uptime % and recent check results
`list_locations`	read	Probe locations for monitor creation
`list_alert_rules`	read	A monitor's alert rules
`list_contact_groups`	read	Notification groups (where alerts go)
`list_incidents` / `get_incident`	read	Incident history and full detail
`list_maintenance_windows`	read	Scheduled maintenance
`create_monitor` / `update_monitor`	write	Create or edit monitors
`delete_monitor`	write	Permanently delete a monitor (irreversible)
`set_monitor_paused`	write	Pause / resume
`run_check_now`	write	Immediate on-demand check
`create_alert_rule` / `update_alert_rule` / `delete_alert_rule`	write	Manage alert rules (catalog-validated)
`acknowledge_incident` / `resolve_incident`	write	Incident actions
`create_maintenance_window` / `delete_maintenance_window`	write	Schedule / cancel maintenance

delete_monitor is irreversible (it also removes the monitor's alert rules and incident history) — the tool description instructs agents to confirm with you first unless you explicitly asked for deletion.

How agents discover your setup

The get_monitor_types tool returns the full monitor-type catalog (target formats, valid alert metrics with operators and example values, config fields), and every validation error names the allowed values — so agents can self-correct instead of guessing. get_summary is the recommended first call: it answers "how is everything right now?" in one round trip.

Notes

The server is stateless Streamable HTTP with JSON responses — no SSE stream, no sessions. Works with every current MCP client.
Tool calls count toward your plan's API rate limit (X-RateLimit-* semantics as in the REST API).
Example prompts: "Any incidents in the last 24 hours? What caused them?", "Create an HTTP monitor for shop.example.com checked every minute from two EU locations", "Schedule maintenance for all monitors tagged backend on Saturday 22:00–24:00 UTC".

MCP Server