MCP tool catalog: everything Claude and ChatGPT can call
The 11 MCP tools, 4 resources, and 5 prompts Lemwatch exposes to AI assistants — with what each one does and who can call it.
Quick answer#
11 tools total — 7 read tools (viewer+), 3 write tools (member+, all require confirm: true), 1 admin-only tool (add_maintenance_window).
Read tools (viewer+)#
list_sites, get_site, get_site_statuslist_incidents, get_incidentget_fleet_health, get_uptime_historyWrite tools (member+, confirm required)#
acknowledge_incidentadd_incident_commentcreate_handoffAdmin-only tool#
add_maintenance_windowResources#
lemwatch://sites — list of monitored siteslemwatch://incidents/active — currently open incidentslemwatch://fleet/health — rolled-up fleet scorelemwatch://sites/{id}/status — one site's live statusPrompts#
incident-triage, root-cause, performance-regression, weekly-client-report, pre-launch-check.
Where to find it in Lemwatch#
Full schemas: /mcp docs page.
FAQ#
Why do writes need confirm?#
Guardrail against prompt-injection attacks — the model must explicitly signal intent.
Can I add custom tools?#
Not yet. Use webhooks for now — see integrations reliability.
Related#
Related Articles
What is the Lemwatch MCP server?
Lemwatch runs a hosted Model Context Protocol server so AI assistants like Claude and ChatGPT can read your sites and act on incidents.
MCP security model and role scopes
MCP tokens inherit the calling member's role, further clamped by an optional role override. Writes require confirm. Admin actions require admin.
How do I create and revoke MCP tokens?
MCP tokens are org-scoped bearer secrets. Create them in Settings, copy once, and revoke any time — hashes are stored, plaintext never is.
Custom Webhook Integration
Send alert data to any HTTP endpoint via custom webhooks. Slack and Discord URLs are auto-routed when used in escalation steps.