Rootly | The Rootly MCP Server

Rootly’s MCP Server is seeing strong adoption across the AI and SRE communities, with nearly 10,000 downloads per month. Today we’re excited to announce it is now generally available, enabling AI assistants and agents to interact directly with your Rootly environment.

Using MCP-compatible tools, teams can query incidents and alerts, automate response workflows, configure platform resources like schedules and workflows, and build AI-driven investigation and remediation workflows.

Why it matters

AI tools are increasingly part of the SRE toolkit, but historically they lacked direct access to live operational systems. The Rootly MCP Server closes that gap by exposing incident operations through a standardized interface that AI assistants and agents can safely interact with.

This unlocks new workflows for reliability teams:

Investigate incidents directly from AI-native environments such as Claude, Codex, or Cursor.
Run AI-assisted root cause analysis using live incident and alert data.
Push suggested fixes or operational changes back into Rootly workflows.
Automate reliability operations through programmable infrastructure and AI agents.

Rootly MCP Server RCA results across 3 months.

Flexible deployment options

The Rootly MCP Server is designed to fit a variety of environments:

Self-hosted deployment for teams that want full control over their infrastructure.
Hosted MCP service for teams that want to get started quickly without managing the server themselves.

Real-time streaming for AI agents

The MCP server now supports both SSE (Server-Sent Events) and HTTP streaming transports, allowing AI clients and agents to stream responses and receive updates in real time using standard HTTP-compatible networking environments.

This improves compatibility with modern AI runtimes, IDE assistants, and enterprise network environments while enabling faster AI-driven incident workflows.

Getting started

Deploy the Rootly MCP Server from the public repository or use the hosted version.
Configure your Rootly API token and connect it to your MCP-compatible client.
Start interacting with Rootly resources from your AI assistant or development environment.

Check out the repo and docs to learn more and get started.

What else shipped

Improvements

On-Call

Functionalities can now be used as paging targets: attach an escalation policy, page via the manual paging modal, and route through workflow actions
When a functionality is paged, the owning team is automatically assigned to the alert
Services can now be added as conditions on escalation policy paths
On-call schedule Slack notifications now include shadow users alongside the primary on-call person
All Slack schedule notifications now show the full user in context
Alert timeline events now render with markdown formatting
Improved voice calling reliability for China numbers

Incident Response

Multiple retrospective exports can now be configured from the same provider
Auto-set configuration is now available for native form fields
The form field placement editor now shows an indicator when a field has auto-set rules configured
The "Checklists" drawer on catalog entity pages now lists all checklists with a detail view
"Start review" now supports selecting from multiple checklists per entity
Checklist review UI has been refreshed with updated row layout, close button, and post-create navigation
Slack command labels have been updated from "Slack Command" to "Manual Trigger" for clarity

API

On-Call

The On-Call API now supports filtering by group_ids

Incident Response

Teams, routes, and other assets can now be referenced by name or slug in API requests, removing the need to look up IDs
API keys can now act on behalf of their associated user, enabling user-scoped API access

Terraform

On-Call

Updated documentation for the delay parameter in rootly_escalation_level
The Opsgenie integration now accepts an empty priority field via Terraform

Incident Response

Fixed an issue where alert routes could not be updated via Terraform
Improved provider retry and backoff behavior when rate-limited during terraform plan

Fixes

On-Call

Alert Slack blocks now post correctly when the "Alert created → Create incident channel" workflow is enabled
HTML form attributes no longer leak into escalation path names when names contain special characters
After deferring and re-evaluating an alert, the system now correctly routes to the urgency-matched escalation path
The escalation policy edit page now warns when navigating away with unsaved changes
Time restriction inputs now display times consistently in 24-hour format
Time restriction UI components are now consistent across escalation path and deferral path editors
Saving an escalation policy with default paging now redirects to the correct page
Deleting an escalation path now shows a confirmation dialog before proceeding
Escalation path conditions no longer allow blank or empty values on creation
Breadcrumbs now truncate long escalation policy names correctly
Urgency option display no longer breaks with very long strings
Tooltips on the escalation policy edit page are now functional
The "Proceed after minutes" field now only accepts valid positive numbers
Users now see a success confirmation after saving escalation policy changes
The time slot dropdown no longer appears as a ghost when "All day" is enabled
Fixed invalid timestamps being accepted by the notification system, causing retry loops and DLQ alerts
The currently selected escalation path is now clearly indicated in the sidebar
Countdown timer now appears for short defer durations, and alert status correctly updates to Triggered
Callers now hear the correct message when a responder is unavailable
Voice call responses are no longer cut off when a responder takes longer to respond
Fixed an RBAC issue where users could delete alerts without the Alerts Write permission
Slack user group on-call tags now update reliably on Sundays
Coverage request UI no longer selects the wrong shift when a time window overlaps another user's shift start
Alerts now auto-resolve correctly in all configurations
Alerts no longer send duplicate Slack messages
Manually creating an alert with "deferred" status now works correctly
Acknowledge and Resolve buttons now appear on deferred alerts in the UI
Deferred alert status now displays correctly on the web
Slack smart defaults no longer fire on deferred alert creation
Preferred names now display correctly on email alerts
Email alert sources now reliably create alerts

Incident Response

Fixed an error when removing a group from sync and attempting to resync
Invalid service account credentials now show a meaningful error instead of silently closing the modal
Long incident titles no longer break the Slack channel topic URL
Fixed an incomplete incident announcement Slack block
Updating a custom form name now correctly updates the linked fields list
Fixed incorrect reporter being set in Jira workflow runs
Fixed a timezone parse error
Integrations and webhooks now correctly receive "deferred" status before "triggered"
Fixed the "IS NOT" workflow operator not being passed correctly in workflow conditions
Status page templates now correctly map template titles to public titles in updates
Fixed the inability to edit the severity native form field in certain cases
The "mark as private" visibility checkbox for form fields now behaves consistently across Slack and web
The payload condition editor is now visible for alert workflows when multiple payload conditions are enabled
Alert workflow Slack messages are no longer overwritten by default context blocks on status changes
Fixed Backstage URLs returning errors on integration pages
Fixed SAML ACS URL mismatch in identity provider configuration
Fixed the public status page SAML endpoint returning 404 for custom domains
Fixed rootly oncall list CLI command returning a 404 error
Slack channel link now appears on the alert block immediately when an alert is created
Fixed Azure AD SSO login returning an error on the public status page