n8n-nodes-referencesource

n8n-nodes-referencesource

n8n community node for Reference Source — verified reference data with a source URL and a verbatim quote on every record. No estimates, no interpolation, no model-generated values: if the source doesn't state it, there's no row.

285 datasets, 96,995 records (checked 2026-09-09): regulatory thresholds, certification and licensing registers, end-of-support and deprecation dates, state-by-state legal variation, version-compatibility tables.

Operations

• List datasets — what's published, how many records each holds, when each was last verified.
• Search records — find records in one dataset by any substring (county, state, chemical, certificate number, product name…).
• Get record — fetch one record by ID.
• Verify quote — check a value you already hold against the published record and its source quote. Returns a verdict, the value the record actually holds, and the source URL either way.

The verdict is one of matches_record (it is the stored value), matches_quote (it isn't the stored value but it is in the verbatim quote — so it probably belongs to a different field of the same source row), differs, or not_found (no such record). It says whether the claim matches what the record and its quote say, not whether the claim is true in the world; the record's source is there so you can go further upstream.

verify_quote is the one this node exists for: hand it a value your workflow already has — from an LLM step, a scrape, a spreadsheet — and it says whether that value matches the published record before anything downstream acts on it.

Example

Search conforming-loan-limits for Autauga → one item per matching record, each carrying source, source_quote, url, last_verified and stale_after.

Verify $806,500 against record 01001 of the same dataset → differs, naming the value the record actually holds ($832,750) and quoting the FHFA row it was read from.

Install

Settings → Community Nodes → Install → n8n-nodes-referencesource

Credentials

The node needs an Apify API token (free account, console.apify.com → Settings → API & Integrations).

The data itself is free and keyless. What needs the token is Apify: this node runs the published Actor referencesource/referencesource-lookup, and Apify bills every Actor run to the account that started it, so a run can't be anonymous. Your token is sent to Apify only; it never reaches referencesource.org.

If you'd rather not involve Apify at all, the same four tools are available keyless as an MCP server at https://referencesource.org/mcp — see referencesource.org/connect. This node exists for workflows, not agents.

Notes

• A search that matches nothing returns no items. Switch on Always Output Data on the node if your workflow needs to see the empty case.
• Every row is stamped with the dataset it came from (dataset_slug, dataset_url, last_verified, stale_after) as well as its own source and source_quote, so a single exported row is still checkable on its own.
• stale_after is the date by which the record should have been re-read. Pages do change: the site re-reads every published quote against its source on a rotation, and rows whose quote has moved are flagged rather than quietly kept.

Links

• Actor: https://apify.com/referencesource/referencesource-lookup
• All datasets: https://referencesource.org/
• Method: https://referencesource.org/method/

MIT licensed.