Skip to main content
This tutorial takes you from nothing to a running workflow. You will scaffold a project, push it, run it on a file, and read the output. Every step is a file or a command, so an agent can do this too.

​
1. Install the CLI

npm install -g @eigenpal/cli
eigenpal auth login
auth login opens a browser to create an API key and stores it locally. One key maps to one tenant.

​
2. Scaffold a project

eigenpal init workflow parse-invoice
cd parse-invoice
You get a self-contained project: workflow.yaml, evaluators.yaml, and a dataset/ folder. Open workflow.yaml to see the step graph.

​
3. Validate and push

eigenpal workflow validate     # check the project locally
eigenpal workflow push         # create the workflow (or a new version)
push prints the workflow id (wf_...). Validation runs locally first, so you catch schema mistakes before anything hits the server.

​
4. Run it on a file

eigenpal run workflows.parse-invoice -F document=@invoice.pdf
File fields use -F name=@path. The command starts a run and prints a run id (exec_...).

​
5. Read the output

eigenpal runs get <run-id>        # status + output JSON
eigenpal runs watch <run-id>      # live, until it finishes
eigenpal runs trace <run-id>      # per-step detail
runs get prints the structured output. runs trace shows what each step did, which is where you look when a step produced the wrong thing.

​
Next