Guides

Task-focused pages. Each shows the agent-led way and the CLI way side by side.

Working with litmus-lite

• Working with an agent: set up Cursor or another MCP client, prompts that work, what a good answer looks like, and the rules the agent follows.
• Reading a report: the HTML report, diagnose, and comparing two runs.
• Running in CI: validate every scenario on each pull request, and run chaos against a live target.

Breaking things

• HTTP faults: latency, error responses and timeouts in front of a service.
• Failpoints: fail one step inside your own Go code.
• Processes and containers: pause or kill a process, pause a container.
• CPU, memory and disk: contention on the machine your service runs on.
• Load and chaos together: run a load generator through the fault, in one scenario.

Going further

• Exporting a scenario: to a neutral JSON form, a LitmusChaos experiment, or a Kubernetes Job.

Walkthroughs

End to end, with a two-process sample app:

• In a terminal: latency, then freezing a dependency process.
• With an agent: the same, prompt by prompt, including consent.