MCP matching reference - Landbase

Matching resolves a company, person, or uploaded dataset to Landbase records. Matching is distinct from enrichment: inspect the match result before starting paid follow-up work.

For AI agents

Use match_company for a domain or company identity and match_person for a person identity.
Use match_dataset for rows in an existing dataset.
Preserve returned IDs, confidence, and account context.
Do not infer a match from an unrelated dataset or run ID.

Match a company

Match acme.example and show the matched company, confidence, and canonical identifiers. Do not enrich it.

{
  "account_id": "5093",
  "domain": "acme.example"
}

Use match_company and inspect the result. For people, provide the strongest available identity fields to match_person. For a dataset, use match_dataset; if it starts asynchronous work, preserve its run identifier and use get_workflow_run.

Match a dataset

Use match_dataset when the input is an existing uploaded or search dataset. The operation creates or returns a matched dataset; it does not change the source dataset in place.

Match dataset ds-7f3c2a. Show the matched row count, confidence fields, and the output dataset ID. Do not enrich it.

{
  "account_id": "5093",
  "dataset_id": "ds-7f3c2a"
}

If the operation returns a workflow_run_id, poll it with get_workflow_run and use the completed output dataset ID for later work. A timeout or lost response does not mean the match stopped; recover the original run before submitting another match.

Match confidence

Treat confidence and candidate identifiers as result data, not as proof that a record is correct. Review low-confidence or ambiguous matches before paid enrichment, contact discovery, or importing records into another system.