Summary

Full Rootly coverage, which means the on-call tools are the ones worth having and the tool count is the thing to manage.

Beyond the expected incident CRUD there is a genuinely useful layer for the person carrying the pager: handoff summaries, shift metrics, responder availability and an override recommendation, plus similar-incident lookup while you are still triaging. The whole surface is far more than a model should carry every turn, so pick the slim profile or an explicit tool list unless you really need the configuration endpoints.

What it is

Rootly's server over its incident management API, covering incidents, alerts, on-call schedules, workflows, catalogs, status pages and the configuration behind them.

What you get

Incidents listed, searched, created and updated, with related incidents found for you and suggested solutions drawn from past ones
On-call intelligence rather than just the roster: shift metrics, a handoff summary, the incidents that landed in a shift, responder availability, an override recommendation, and a health-risk check on the rotation
Alerting configured end to end — sources, alerts and alert events, groups, routes and routing rules, urgencies, and alerts attached to an incident
Schedules and rotations built from the parts: rotations, their users and active days, override shifts, on-call roles and shadows, and escalation policies with their levels and paths
Post-incident work: action items, retrospective processes and their steps, timeline events, and meeting recordings
Workflows created and run, with their tasks, groups and form-field conditions, plus playbooks and their tasks
The service catalog and the objects incidents hang off — services, teams, functionalities, environments, severities, incident types, sub-statuses and causes
Custom fields and forms, form sets and field placements, so incident intake matches how your organisation actually reports
Status pages and their templates, heartbeats, dashboards and panels, and uptime and incident charts per service, functionality or team

Requirements

A Rootly account. The hosted server needs nothing installed: point the client at https://mcp.rootly.com/mcp over streamable HTTP and complete OAuth2 in the browser, or send a Rootly API token as a bearer header; SSE and a code-mode endpoint are available too. To run it yourself, install rootly-mcp-server from PyPI with ROOTLY_API_TOKEN. The full tool surface is large — the hosted slim profile, selected with a tool_profile parameter or the X-Rootly-Tool-Profile header, cuts it to about 70 tools, and ROOTLY_MCP_ENABLED_TOOLS sets an exact list.

Setup effort

One command plus a key — claude mcp add --transport http rootly https://mcp.rootly.com/mcp, then supply credentials