Chaos Toolkit vs LitmusChaos

Chaos Toolkit is a free, open-source chaos engineering toolkit for developers exploring systems to uncover weaknesses. Experiments define a steady-state hypothesis and probes to check it, along with a method of actions and probes. If the system does not meet the stated steady state, the method is not applied and the experiment must stop. Experiments can be declared in JSON or YAML files, making them suitable for collaboration and orchestration alongside code. The toolkit extends to systems through its Open API and can be embedded in a CI/CD chain. Listed integrations include AWS, Azure, Google Cloud Platform, Kubernetes, Kafka, Slack, Datadog, Grafana, OpenTelemetry, and Prometheus. The standalone CLI is available under Apache License 2.0. It is implemented in Python 3, officially supports Python 3.8 and later, and has only been tested against CPython. Installation instructions cover macOS, Debian or Ubuntu, and Windows; users can install extensions after the CLI and core library. Reliably is a separate platform with SaaS, private SaaS, and on-premises options.

LitmusChaos is an open-source chaos engineering platform for finding infrastructure weaknesses through controlled tests. It is designed for developers, Kubernetes developers, and SREs, and supports Kubernetes along with cloud and network fault injection. Experiments can be declared, scheduled, browsed, and analyzed, and can run in sequence or in parallel. ChaosHub provides tested experiments that users can tune to get started. Probes let teams define and check steady-state hypotheses within chaos scenarios. Litmus exports Prometheus metrics so teams can assess impact through in-house dashboards or external visualization and APM tools. The platform can be installed on a user's own cluster, with Helm offered as an installation option, or used through hosted SaaS; its deployment model is hybrid. A Backstage plugin displays experiments, Chaos Hubs, environments, and experiment runs. The Litmus MCP Server works with Model Context Protocol clients, including Claude Desktop, to discover, run, and observe experiments, using tokens and scoped access. The project is CNCF-hosted and identifies itself as a CNCF Incubating Project. The hosted SaaS and open-source platform plans are both listed at free.