File to JSON Converter (CSV, Excel, XML, iCal, vCard, GPX)

AI agents, APIs and Apify Actors work with JSON, but people have spreadsheets, exports and attachments. Give this Actor a file (a URL, the text, or base64) and it returns clean JSON records, one per row, event, contact or point. It works out the format itself. It's the reverse of dataset-format-converter, and every format that writes is read back here in the tests.

Formats

format	Reads	Notes
csv, tsv	Any CSV style 1: comma, semicolon (European Excel), tab or pipe; BOM; quoted line breaks; UTF-8, UTF-16 or Windows-1252	Delimiter detected from the first line
xlsx	Excel workbooks 2	Pick a sheet by name or number. Numbers, true/false and dates come back typed; dates are found from the cell's number format. Formulas give their saved value; error cells become null, with a warning. Old .xls isn't supported
xml	Any XML 3	Records are the most repeated element under the root (or recordElement). Attributes become @name, repeated children become lists. DOCTYPEs are refused, which blocks entity-expansion attacks 8
ics	iCalendar events and to-dos 4	Folded lines joined, text unescaped. All-day dates as YYYY-MM-DD, UTC times with Z, local times with their timeZone
vcard	vCard 3 and 4 contacts 5	fn, name, org, title, email, tel, url, address, note
gpx	Waypoints, routes and tracks 6	lat, lon, elevation, time, name, with track and segment numbers
geojson	Feature collections 7	Properties plus lat/lon for points, or the geometry as JSON
json, jsonl	Lists, wrapped lists ({"data": {"results": [...]}}), JSON Lines	The list of records is found automatically (or set recordPath)
fixed-width	Mainframe and legacy text files	Needs a layout: the one dataset-format-converter writes, or a list of widths

Values that are clearly numbers or true/false are typed. Text with leading zeros, such as account numbers, phone numbers and 00123, stays text. Decimal commas stay text too, because 1,234 is ambiguous. Set typeValues: false to keep everything as text.

Input

{
  "text": "Site;Region;Capacity MW;Live;Account\nExample site A;West Midlands;12.5;true;00123\n\"Example site B; phase 2\";East of England;40;false;00456\n",
  "fileName": "sites.csv"
}

Give exactly one of url (public http or https links only; local and private-network addresses are refused 9), text or base64. Files can be up to 20 MB and give up to 100,000 records.

Example output

The first record:

{
  "Site": "Example site A",
  "Region": "West Midlands",
  "Capacity MW": 12.5,
  "Live": true,
  "Account": "00123"
}

The dataset holds one item per record. The OUTPUT record adds the format found, the columns, details (delimiter, sheet names, record element) and warnings.

Integrations

Ready-made files in integrations/README.md: a Microsoft Power Platform custom connector (Power Automate, Power Apps, Logic Apps, Copilot Studio), OpenAPI 3 (Azure API Management, Postman, n8n, AI agents), an Excel / Power BI Power Query script, a Google Sheets function, and curl, Python and JavaScript examples. Apify's own Zapier, Make and n8n apps and its MCP server work too, so an AI agent can hand it an attachment link and get records back.

Pricing

Pay per event: $0.30 per 1,000 results ($0.0003 each), plus Apify's minimal start fee. A result is one row in the dataset: one record read from your file. The run stops at your spending limit.

Sources

1. RFC 4180: Common Format and MIME Type for CSV Files. Accessed 3 October 2026.
2. Ecma International: ECMA-376 Office Open XML File Formats. Accessed 3 October 2026.
3. W3C: Extensible Markup Language (XML) 1.0. Accessed 3 October 2026.
4. RFC 5545: iCalendar. Accessed 3 October 2026.
5. RFC 6350: vCard Format Specification. Accessed 3 October 2026.
6. GPX 1.1 Schema Documentation. Accessed 3 October 2026.
7. RFC 7946: The GeoJSON Format. Accessed 3 October 2026.
8. OWASP: XML External Entity Prevention Cheat Sheet. Accessed 3 October 2026.
9. OWASP: Server-Side Request Forgery Prevention Cheat Sheet. Accessed 3 October 2026.