Export Notion, HubSpot & Airtable Records to a Dataset

Your records are stuck inside an app. This Actor pulls them out into an Apify dataset, where the rest of your pipeline can reach them — enrich them, scrape against them, analyse them, or just download a clean CSV.

Point it at a connected app, name the tool that lists records, and run.

What it handles that a manual export doesn't

• Pagination. Give it the cursor argument and where the next cursor appears, and it walks every page. A repeated cursor stops paging immediately, so a misbehaving server can't loop.
• Incremental exports. With onlyNew, a scheduled run returns only records that appeared since last time — not the whole table again.
• Flattening. Nested objects become dot-notation columns (props.name), so the export opens cleanly as a spreadsheet instead of a wall of JSON. Arrays stay intact.
• Field selection. Export the six columns you need, not all forty.

How it works

1. Connect your app under Integrations in Apify Console.
2. Run in list-tools mode to learn the connector's tool names.
3. Run in pull mode.

{
  "source": "<your connector>",
  "mode": "pull",
  "readTool": "query_database",
  "readToolArgs": { "database_id": "abc123" },
  "cursorArgName": "start_cursor",
  "cursorPath": "next_cursor",
  "idField": "id",
  "onlyNew": true,
  "flatten": true
}

Credentials never reach this Actor. It talks to your app through Apify's MCP proxy, which holds the real tokens.

Examples

Notion database → CSV — readTool: "query_database", flatten: true. Download the dataset as CSV when it finishes.

HubSpot contacts → enrichment pipeline — pull contacts, then feed the dataset into an enrichment Actor and write the results back with the Field Filler.

Nightly incremental Supabase export — readTool: "select", idField: "id", onlyNew: true, on a schedule. Each run returns only rows added since the last one.

Input

Field	Required	Description
source	yes	MCP connector to export from
mode		pull (default) or list-tools
readTool	for pull	Tool that lists records
readToolArgs		JSON args for the read tool
cursorArgName		Paging argument the tool accepts, e.g. start_cursor
cursorPath		Where the next cursor appears in the response. Dot paths work
maxPages		Hard stop on pagination. Default 10
idField		Field uniquely identifying a record
onlyNew		Only export records unseen in previous runs. Needs idField
fields		Allow-list of fields to keep. Empty means all
flatten		Nested objects → dot-notation columns
maxItems		Cap on records exported. Default 5000

Output

Records land in the dataset in their own shape — exactly the fields the source returned, optionally flattened and filtered. Export as JSON, CSV, Excel or XML, or read it from the API.

Pricing

$5 per 1,000 records exported. With onlyNew on, a scheduled run over an unchanged source exports nothing and costs nothing.

FAQ

Does this see my Notion/HubSpot token? No. Apify's MCP proxy holds credentials and attaches them to outbound calls.

Which apps work? Any service exposed as an MCP connector in your Apify account.

How do I find cursorPath? Run once without pagination and look at the raw result, or use list-tools to read the tool's own schema.

Does it modify anything in the source app? No. This Actor only reads.

Can an AI agent call this? Yes — it's exposed through the Apify MCP server, and it's designed to return a bounded number of records rather than an unbounded dump.

---

Need the rest of the record on these companies?

This actor returns one source. KYB Company Data Enrichment takes a company name and returns what the public record says about it across up to nine sources — state business registries, UCC lien filings, tax warrants and WARN layoff notices, the FMCSA motor-carrier census, the CMS NPI Registry, USAspending federal awards, EPA ECHO compliance, SEC EDGAR filings and CourtListener federal dockets.

Every enriched field carries its source and its match confidence. Only matches above the accept threshold are merged — anything weaker is surfaced as a suggestion and never folded into the record, because a false join is worse than returning nothing.