Execution Workflows

Munk AI supports multiple execution paths. The right command depends on how much structure you already have: are you running a single known case, executing a reviewed plan, or validating a real code change?

1. Structured Execution (run case)

Use munk run case when you have a specific, structured TestCase you want to execute. This is the most predictable single-case path because the AI follows predefined intents and expectations.

Best for:

• Running a test you previously recorded via the Web UI.
• Validating a specific, known scenario repeatedly.

Quick Example (Web)

munk run case \
  --app-id web-demo \
  --plan-id my-first-plan \
  --case-id login-success-case \
  --platform web \
  --base-url https://example.com \
  --headed \
  --config ./munk.yaml

Tip: The --headed flag opens a visible browser window so you can watch Munk AI perform the actions.

2. Batch Execution (run plan)

Use munk run plan when you want to execute a collection of related cases together. Instead of passing arguments via CLI flags, you provide a structured JSON request file containing multiple cases.

Best for:

• CI/CD pipelines.
• Running a full regression test suite.

Quick Example

munk run plan \
  --request-file ./my-test-plan.json \
  --config ./munk.yaml \
  --json

Tip: Adding the --json flag tells Munk AI to output machine-readable results instead of human-readable text, which is perfect for parsing in CI environments.

3. Change-aware Verification (verify change)

Use munk verify change when you want Munk AI to reason from a code change, generate a verification plan, and optionally continue into execution.

Best for:

• validating a real implementation change before merge
• generating focused regression coverage from change context
• agent-driven or developer-driven review-first workflows

Quick Example

munk verify change \
  --app-id app-1 \
  --change-summary "Fix task save flow" \
  --changed-file src/task.py \
  --config ./munk.yaml

4. Multi-plan batch runs and scheduled tasks (Web UI)

When you need to run multiple plans serially on one device, or turn the same setup into a recurring regression task, use the Web UI batch run screen instead of the CLI.

Best for:

• Nightly or pre-release multi-plan serial regression
• Automatically triggering the same plan set on a fixed cadence
• Submitting several plans at once after maintaining them in Tests

See Batch Runs and Scheduled Tasks for the full walkthrough.

Summary: Which one to choose?

• Testing a single, concrete scenario? --> munk run case
• Running a reviewed suite of cases? --> munk run plan
• Running multiple plans at once in the Web UI, or creating a scheduled task? --> Batch Runs and Scheduled Tasks
• Validating a code change end-to-end? --> munk verify change

What's Next?

After a successful execution, you might want to:

• Learn how to use these executions to validate actual PRs: Verify Code Changes.
• Avoid writing JSON cases by hand: Record & Replay.
• Learn multi-plan batch runs and cron scheduling in the Web UI: Batch Runs and Scheduled Tasks.