Skip to main content
Version: Next

Example Interactions

In this documentation, you can find copy-paste-ready interactions you can perform via the MCP server. Each example pairs a natural language prompt with the underlying tool calls and a sample response so you can replicate the workflow quickly.

If you’re new to the tool surface, see mcp-server/available-tools.md for a complete reference of capabilities and parameters.

Quick Start​

A common end-to-end flow looks like this:

  1. List experiments → pick one to run.
  2. Run an experiment → capture the Run ID.
  3. Monitor the run → check steps, logs, and probe results.
  4. (Optional) Stop the run if needed.

Below, you’ll find detailed examples for these and more scenarios.

Sample Prompts​

  • Prompt

    Show me available chaos experiments in the staging environment that target Kubernetes.

    It will: List chaos experiments filtered by environment and platform so you can choose one to run.

  • Prompt

    Run experiment "pod-delete-basic" now in staging and return the run ID.

    It will: Trigger an on-demand run of the chosen experiment and return the Run ID.

  • Prompt

    Show me the latest status, timeline, and probe results for run ID <RUN_ID>.

    It will: Retrieve detailed run information including step timeline and probe outcomes.

  • Prompt

    Stop the currently running run <RUN_ID> and confirm the termination.

    It will: Attempt to stop the in-progress run and report acceptance.

  • Prompt

    List all registered infrastructures and show which ones are healthy.

    It will: Return the registered infrastructures with their connection/health status.

  • Prompt

    Onboard a new Kubernetes cluster named "edge-lab" and provide the registration steps.

    It will: Initiate registration and return the manifest or token with next steps.

  • Prompt

    Create an HTTP probe that checks GET https://myapp.example.com/health returns 200 in under 2s.

    It will: Create a reusable HTTP probe definition for steady-state validation.

  • Prompt

    List all probes so I can attach one to my next experiment.

    It will: List available resilience probes with IDs and brief specs.

  • Prompt

    Show me available ChaosHubs and then list Kubernetes pod-level faults.

    It will: List connected hubs and then fetch faults filtered by platform/category.

  • Prompt

    Create a new environment called "chaos-lab" for ad-hoc testing.

    It will: Create a new environment grouping that you can target in experiments.

  • Prompt

    List environments so I can verify "chaos-lab" exists.

    It will: List all environments with their IDs and names.

Tips and Good Practices​

  • Start broad with listing tools (list_chaos_experiments, list_experiment_runs, list_chaos_infrastructures, list_environments) before drilling down.
  • Prefer IDs over names for precision when running or stopping experiments.
  • After run_chaos_experiment, immediately capture the runId to monitor or stop it later.
  • Reuse probes across experiments to standardize resilience checks.
  • Keep filters small and focused to reduce noise in large projects.

For detailed parameter schemas and additional examples, see mcp-server/available-tools.md.

Learn more​