How to enrich a dataset with MCP
Use MCP dataset workflows when you want to add fields to many rows, preserve dataset lineage, or continue processing the result in another workflow. This guide covers the recommended sequence for enrichment and related dataset operations.Choose the right operation
Use
transform_dataset first when you need to reduce the number of rows processed by a paid workflow. Use enrich_dataset for structured fields, qualify_dataset for a yes/no or scored decision, and research_dataset for generated findings.
1. Inspect the dataset
Start by checking the dataset fields instead of assuming that every dataset has the same schema. Ask your MCP client:
Show the row count and available fields for dataset ds-7f3c2a. Do not start any paid work.
Use get_dataset and get_dataset_fields to inspect the dataset. Confirm that it contains a reliable company, person, or contact anchor for the operation you want to run.
2. Limit the input when needed
If the dataset is larger than necessary, create a smaller working dataset before enrichment:
Create a read-only working dataset containing the first 100 rows from ds-7f3c2a. Show the resulting row count and fields. Do not enrich it yet.
Use transform_dataset with an explicit, read-only filter or limit. Keep the returned output_dataset_id; it becomes the input for the next step.
3. Start the enrichment
Ask the client for the fields you want:For the working dataset, add company industry and employee count.The client calls
enrich_dataset with confirm: true, the selected
account_id, and the exact field names returned by get_dataset_fields.
Landbase deducts any credits when the work runs; the client does not stop to
ask you to approve the cost.
4. Optional: see the cost first
For qualification, research, or contact enrichment, you can review the cost before the work starts:Check whether each company in the working dataset sells to hospitals. Show me the credit quote first and wait for my answer.The client calls the tool without
confirm, shows the returned
needs_confirmation quote, and waits. When you agree, it repeats the same
request with confirm: true. Do not change the dataset, fields, account, or
row scope between the quote and the confirmed call without requesting a new
quote.
5. Poll the workflow
Enrichment is asynchronous. Keep the returnedworkflow_run_id and poll it with get_workflow_run until it reaches a terminal state.
If a polling call times out, the workflow may still be running. Continue with the same run ID instead of starting a duplicate enrichment. Keep the same account_id on every status call; see Continue a search or workflow.
6. Continue with the output dataset
When the workflow succeeds, use itsoutput_dataset_id for the next preview, workflow, or download. Enrichment creates a child dataset; it does not silently replace the source dataset.
Ask your MCP client:
Show the fields and row count for the enriched output dataset.Use MCP enrichment reference for the exact request fields and MCP workflow reference for workflow status and output details.
Contact enrichment
Useenrich_contacts when the input contains known contacts and the goal is discovering work email or phone data. Inspect the input first and preserve the returned identifier for recovery. See MCP contact enrichment reference.
For company or person attributes on an existing dataset, prefer enrich_dataset instead of contact enrichment.

