LitmusChaos is an open-source platform for identifying infrastructure weaknesses with controlled chaos tests. It is designed for developers, Kubernetes developers, and SREs, and supports Kubernetes as well as cloud and network fault injection. Experiments can be declared, scheduled, browsed, and run in sequence or in parallel, with analytics to review. ChaosHub provides tested experiments users can tune as a starting point, while probes let teams define and verify steady-state assumptions within scenarios. Litmus exports Prometheus metrics for assessing impact in in-house dashboards or external visualization and APM tools. It can be installed on a user's own cluster, with Helm offered as an installation option, or used as hosted SaaS. The hybrid deployment model includes blast-radius controls. A Backstage plugin displays experiments, Chaos Hubs, environments, and runs. Its MCP Server works with compatible Model Context Protocol clients, including Claude Desktop, to discover, run, and observe experiments; the homepage says it uses tokens and scoped access. Litmus identifies itself as a CNCF-hosted Incubating Project.
Who it is for
LitmusChaos suits developers, Kubernetes developers, and SREs who want to run controlled infrastructure experiments. It may also fit teams seeking cluster installation or hosted SaaS, with Prometheus metrics and Backstage integration.
What is good
- Experiments can run in sequence or in parallel.
- ChaosHub offers tested, tunable experiments.
- Prometheus metrics support impact assessment.
- Helm is an installation option.
- Blast-radius controls are available.
What to know first
- Self-hosted use requires installation on a user's cluster.
- MCP Server requires a compatible Model Context Protocol client.
Verdict
LitmusChaos provides experiment authoring, scheduling, probes, and analytics for controlled chaos testing. Teams can choose cluster installation or hosted SaaS and connect experiment results to their existing observability tools.
LitmusChaos plans and pricing
All plansCompared on chaos engineering platforms
- Kubernetes support
- Yes
- Cloud fault injection
- Yes
- Network fault injection
- Yes
- Experiment scheduling
- Yes
- Deployment model
- hybrid
- Blast-radius controls
- Yes
