celesto
Secure and persistent computer for AI agents
Correlate observability signals and Kubernetes resources
Translation pending
Developer toolsGitHub @korrel8rOnlineOwner unconfirmed
Have you tried it? One tap is enough. No sign-in needed.
Repository information has not been collected yet.
No updates detected yet. GitHub releases will appear here.
Correlation of observability signals.
Translation pending
Loading the full README.
Read the full README ↗Korrel8r is a rule-based correlation engine for Kubernetes clusters.
It connects cluster resources and observability signals — logs, metrics, alerts, traces, network flows — into a graph of related data. It searches multiple back-ends with different APIs and query languages automatically. Given any starting point — an alert, a pod, a deployment — it finds related data across multiple signal stores, such as Prometheus, Loki, Tempo, and the Kubernetes API.
Neighborhood Search : Find everything reachable within N steps. "Show me everything related to this pod" — returns owning deployments, logs, metrics, network flows, and more.
Goal Search : Find paths to a specific type of data. "Find all logs related to this alert" — korrel8r connects the alert to deployments, deployments to pods, pods to logs.
Interactive troubleshooting : The OpenShift troubleshooting panel displays interactive correlation graphs using Korrel8r as a back-end. Click to jump to relevant resources directly, without manually discovering the intermediate relationships.
AI-agent troubleshooting : Korrel8r provides an MCP interface for AI agents. Its rule-based engine searches and correlates large data sets faster than an AI agent can. It builds a small, focused graph of relevant signals, so the AI can concentrate on understanding the problem rather than spending tokens discovering connections.
Custom automation : Korrel8r provides REST and MCP interfaces, can run as a cluster service or a command, and can be used as a Go library. Embed it in custom automation solutions that need correlation data.
Rules : Take data from a "start" class to generate a "goal" query for related data. Rules are written as YAML files using Go template syntax.
Domains : A "domain" defines a signal type, query language, and data store. New domains can be added to the engine. Coding is required.
# Latest release
go install github.com/korrel8r/korrel8r/cmd/korrel8r@latest # Executable
docker pull quay.io/korrel8r/korrel8r:latest # Image
# Daily images from main for development, not tested for release.
docker pull quay.io/korrel8r/korrel8r:dev-latest
See all available images at: https://quay.io/repository/korrel8r/korrel8r?tab=tags
We welcome contributions! See CONTRIBUTING.md for how to get started.
Korrel8r includes custom commands for Claude Code to streamline development:
/generate-rule - Interactive assistant for creating correlation rules.This project is licensed under the Apache License 2.0 - see the LICENSE file for details.