Run your first blueprint

For developers with the CLI installed and a reviewed local blueprint package. This tutorial covers preflight, launch, inspection, and cancellation; it does not install a particular catalog or promise a domain-correct result.

Before you begin

Complete Installation. Obtain a blueprint folder from a catalog or its author, and read its README, referenced execution/configuration documents, dependencies, and payloads. Use its documented sample inputs. Commands below run from the directory containing that folder; replace ./my-blueprint with its actual relative or absolute path.

Review worker code, mounted files, passed environment variables, model endpoints, and external actions before launch. Validation does not establish that a package is safe to execute.

Step 1: Start and inspect the runtime

mn runtime start
mn runtime status

Resolve failed required components before submitting work. Use mn runtime doctor for deeper diagnostics. Startup prints a federation join credential; keep it private.

Step 2: Validate and diagnose the blueprint

mn blueprint validate ./my-blueprint
mn blueprint doctor ./my-blueprint

Validation checks the local package and declared requirements. Doctor diagnoses launch prerequisites. Resolve reported models, services, inputs, or hardware requirements before continuing; see Model Runtime and Troubleshooting. There is no universal model requirement for all blueprints.

Local launch and doctor targets must begin with ./, ../, or /; other targets are interpreted as catalog IDs. Use Examples to select a catalog entry.

Step 3: Launch

mn blueprint run ./my-blueprint --detached

Record the returned <job-id> and <run-id>. The job is a durable definition; the run is this execution. Starting a run may prepare dependencies and runtime resources and execute the blueprint's external actions. --detached skips the live workflow UI; it does not cancel the execution.

Step 4: Inspect the result

mn job show <job-id>
mn run show <run-id>
mn run watch <run-id>
mn run logs <run-id> --channel logs
mn run logs <run-id> --channel events
mn run result <run-id>

Ctrl+C detaches from the watcher. Inspect the terminal run state, warnings, artifacts, and any required human review before using the result. Completion means execution finished, not that its domain conclusions are correct.

run result downloads outputs into $MN_HOME/outputs/<run-id> by default. Blueprint run records live under $MN_HOME/runs/<run-id>, with MN_HOME defaulting to ~/.mn. For pending human requests and report export, see Monitor.

Clean up

If the run is unfinished and should stop:

mn run cancel <run-id>

Cancellation cannot undo external actions already performed. Preserve diagnostic records and outputs. Stop local services when other runs no longer need them:

mn runtime stop

Next steps

• Examples: select another blueprint.
• CLI Reference: create reusable jobs, start runs, and inspect output.
• Blueprint Standard: author a package.
• Security Model: review execution and data boundaries.