Skip to main content
Use this guide when you know your starting data and your desired result, but not which Play to use. Pick one outcome, inspect the matching Play, then pilot it on a small, representative input.

Before you choose

Have these three things ready:
  • The data you can start from, such as a CSV with names and domains, a list of account domains, or a visitor event.
  • One outcome, such as a verified work email, a researched account, or a reviewed ads audience.
  • A small representative sample. Do not use a full list to discover whether a workflow fits.

Match your input to the job

Inspect a candidate without running it

Use search to find a maintained Play, then read its input contract. These commands do not start a run or spend Deepline credits.
plays check confirms that the current prebuilt contract is available. It does not deploy the Play or call a provider. Use a prebuilt Play for one bounded lookup with a maintained route and result rule. Use a workspace-owned Play when you need code, branching, approval logic, a trigger, or an external write. For a name-and-company lookup that should return a LinkedIn profile, start with prebuilt/person-to-linkedin-harvestapi. The same name without -harvestapi remains available for compatibility with existing saved workflows.
Continue only when the Play accepts the fields you have and returns the result you need. If either is unclear, ask your agent to explain the contract before it runs a pilot.

Ask an agent to choose when the table does not fit

Give your coding agent the file name, its column names, the outcome, and the pilot boundary:
Choosing a Play is not approval to run paid work. Keep the pilot and approval boundary explicit.

Next step

Open the guide for the job you selected. Each guide explains its prerequisites, safe validation step, pilot, and the result to review before you scale.