Agentic AI Platform Change Log

Human approval / guardrails

AI Instructions are now available in every Rootly organization: admins set a global instruction plus instructions for each feature that the Rootly Agent, incident summaries, AI in Retrospectives and Proactive Intelligence follow. Rootly maintained best practice instructions are on by default, can be switched to custom, and every change is kept in an edit history with a line by line diff.

Teams can keep the agent from putting customer or internal service names into status page text and keep retrospectives blameless without writing prompts per feature, and the admin only control and diff history record who changed agent behavior.

Workflow orchestration

Rootly added workflow actions that create Slack Canvases and update them as incident status, severity, roles, and tasks change. Markdown templates and Liquid variables populate incident data, while managed sections preserve notes written by responders. Customers must reconnect Slack to grant the new canvas permissions.

Incident stakeholders can follow a current summary without interrupting responders for manual updates. Preserving human notes supports collaborative incident records alongside automated content.

Agent capability

Rootly has expanded its AI agent to its mobile application, allowing responders to manage incidents via chat on the go. The mobile agent can generate catch-up briefings, draft customer-ready status updates, and execute incident actions like paging or changing severity.

On-call engineers can now orchestrate incident responses directly from their phones without needing to open a laptop. The system enforces human-in-the-loop guardrails by requiring explicit confirmation for high-stakes actions to maintain operational safety and auditability.

Integrations

Rootly released the Edge Connector, enabling Rootly AI to connect directly to users' existing observability stacks. This integration allows the platform to go from sign-up to incident-ready in minutes by seamlessly ingesting observability data.

Engineering teams can now rapidly deploy Rootly's AI capabilities by plugging it straight into their current monitoring tools. This drastically reduces onboarding time and ensures the AI agent has immediate access to critical telemetry during incidents.

MCP / tool calling / API

Rootly introduced the Rootly Wizard, a guided command-line interface tool accessed via an npx command that automates initial workspace configuration. The tool allows administrators to instantly set up teams, on-call schedules, alert routing, Slack integrations, and status pages, culminating in a live end-to-end test page.

This drastically reduces time-to-value and onboarding friction for new deployments. Engineering teams can validate their entire incident configuration, including paging a real device, before broadly rolling the platform out to users.