Kept, developed by Egroup Labs, is an open-source AIOps and alert management platform designed for the mcp platform. Built primarily with Python and TypeScript, it offers solutions for managing alerts in complex operational environments. Kept is distributed via Docker images, Docker Compose, and supports deployment on Kubernetes, AWS ECS, and OpenShift. This free platform helps SREs, operators, and engineers automate responses and enrich incident data across diverse systems.

What does Kept connect to?

Kept integrates with over 110 external systems, functioning as an AIOps and alert management platform. It connects to various monitoring tools, incident response management systems, ticketing systems, source control, change management, and configuration management databases. Specific integrations include CloudWatch, Sentry, Datadog, Prometheus, PagerDuty, and Jira. This allows it to centralize alert processing and data from many operational tools.

Core features and tools

Kept's workflow engine uses Triggers, Steps, and Actions. Triggers initiate workflows based on alerts, incidents, schedules, or manual input. Steps read or fetch data for enrichment and context. Actions execute operations like updating tickets, sending notifications, or restarting servers. It also integrates with external systems through providers. Specific actions include querying MySQL databases, enriching alerts with query results, updating Jira tickets, and executing Python scripts.

Who is it for?

Kept is for SREs, operators, engineers, startups, and global enterprises that manage alerts in complex environments. It helps teams automate responses and gain context from a variety of sources. This platform is particularly useful for those needing to orchestrate workflows across many operational tools. This is not the right fit if you require explicit confirmation of specific MCP server transport protocols or compatible AI clients.

How to get started

Access is granted through a standard OAuth authorization flow, with support for Auth0, Keycloak, Okta, OneLogin, and Azure AD. Alternatively, DB authentication or No Authentication are options, with the authentication scheme controlled by environment variables such as AUTH_TYPE, KEYCLOAK_ID, and KEYCLOAK_ISSUER on both the backend and frontend components.

Use cases

• Automatically query Kubernetes via API to check for deployments and bundle context for alerts from monitoring tools like Datadog, only alerting a team if a specific condition (e.g., pod not restarted) is met.
• Auto-create tickets, send Slack notifications, or enrich alerts with external data from various sources.
• Receive alerts, query a MySQL database based on alert content, and send webhooks with the query results to other systems.

Kept offers a versatile platform for managing and automating responses to operational alerts. Its ability to integrate with over 110 services and orchestrate complex workflows makes it a valuable tool for SREs and engineers. The platform supports various authentication models, providing flexibility for different organizational security needs. Start by configuring an authentication method like Keycloak and setting up a workflow to query a MySQL database for alert enrichment.