Skip to content

Set up and administer a project

As an administrator, you decide what labelers read, what they answer, and what the finished dataset contains. You describe those decisions in a .jsonc configuration file. A schema is simply the definition of those fields: their names, value types, and rules.

Start with the question you need answered. These are actual MLabel controls; each link shows the configuration that creates it.

See all controls, their settings, and validation states →

The illustrated cookbook pairs every recipe with screenshots, a downloadable configuration, and sample data. Choose a recipe for multiple answers, reviewer details, source tables, warning rules, or another common task.

Create a small working project →

The tutorial includes a complete configuration and sample CSV. You will copy a record ID, ask for a decision, add optional notes, and check the export.

  1. Plan the output: define what each column means and which answers are required.
  2. Choose field types and controls: see what labelers will actually use.
  3. Add clear captions and help and arrange fields into cards.
  4. Validate the configuration, then pilot it on representative records.
  5. Distribute the app and configuration and hand the data to a preparer.
Question Guide
What belongs in a configuration file? Configuration structure
Which values can a field hold? Value types
Where does an output value come from? Copy, session, user, and timestamp fields
How do I highlight information for a labeler? Display rules
How do project shortcuts work? Configuration shortcuts
How do I disable update traffic? Network policy
What does this validation error mean? Error reference
What are all the allowed keys? Generated schema reference

Use a new configuration filename when changing a live project’s schema, and keep outputs from different versions separate until you have checked compatibility.