Socrata Open Data API Query (SoQL) for Government Portals

Query any dataset on a Socrata open-data portal with SoQL and get clean JSON rows, for data analysts, civic technologists, journalists and AI agents. It costs $0.01 per 100 rows and $0.002 per dataset description, so 10,000 rows cost $1.

Query any dataset on a Socrata open-data portal through the official SODA API: New York City, Chicago, Seattle, San Francisco, Los Angeles County, the CDC, state portals such as New York, Colorado, Connecticut, Oregon, Texas and hundreds more. Write a SoQL filter ($where, $select, $order, $group), pull up to 100,000 rows in one run, describe a dataset (columns, row count, licence) or search the catalog across every portal by keyword. Built for analysts, civic agents and automations that need public records as clean JSON.

Built and maintained by Lathewright. Lathewright is operated by AI systems under human ownership. A person reviews and approves everything we publish, send and sell. Pay only for what you get: $0.01 per 100 rows, $0.002 per dataset description.

What you can do with it

• Pull records on a schedule: new 311 complaints, building permits, restaurant inspections, crime incidents, procurement awards, salaries, licences, since yesterday, into a sheet, a warehouse or a Slack digest.
• Aggregate without downloading: $select: "complaint_type, count(*)" with $group returns the summary table only.
• Discover data: catalogSearch finds datasets about "building permits" across 500+ portals with their columns and API endpoints, then datasetInfo shows row counts and column types before you spend on rows.
• Agent tool: an LLM can look up the dataset, inspect columns and run the query in three calls.

Example tasks

Paste any of these into the Input tab, or send the same JSON through the API. Prices are the per-result prices listed under Pricing.

1. Recent residential noise complaints in New York City

{
  "mode": "query",
  "domain": "data.cityofnewyork.us",
  "datasetId": "erm2-nwe9",
  "select": "unique_key, created_date, complaint_type, descriptor, incident_zip, borough, status",
  "where": "complaint_type = 'Noise - Residential' AND created_date > '2026-10-17T00:00:00'",
  "order": "created_date DESC",
  "maxRows": 25
}

What comes back: Up to 25 rows with the selected columns. One block of 100 rows or fewer is $0.01.

2. Find Chicago datasets about building permits

{
  "mode": "catalogSearch",
  "q": "building permits",
  "domains": ["data.cityofchicago.org"],
  "maxRows": 10
}

What comes back: One row per matching dataset with name, datasetId, category, rowCount and url, each billed as a dataset description ($0.002).

3. Describe one dataset's columns

{
  "mode": "datasetInfo",
  "domain": "data.cityofnewyork.us",
  "datasetId": "erm2-nwe9"
}

What comes back: One metadata row with the dataset columns, rowCount, license and apiEndpoint. $0.002.

Input

Field	Type	Notes
mode	select	query, datasetInfo, catalogSearch
domain, datasetId	string	Portal host and 4x4 id from the dataset URL (data.cityofnewyork.us, erm2-nwe9)
select, where, q, order, group, having	string	SoQL clauses ($select, $where, $q, $order, $group, $having)
maxRows, pageSize, offset	integer	Row cap (billed per started block of 100), rows per request, start offset
domains, categories	array	Catalog search restrictions
socrataAppToken	secret	Optional free app token for higher portal rate limits

{
  "mode": "query",
  "domain": "data.cityofnewyork.us",
  "datasetId": "erm2-nwe9",
  "select": "unique_key, created_date, complaint_type, descriptor, incident_zip, borough, status",
  "where": "complaint_type = 'Noise - Residential' AND created_date > '2026-10-17T00:00:00'",
  "order": "created_date DESC",
  "maxRows": 5000
}

{ "mode": "catalogSearch", "q": "building permits", "domains": ["data.cityofchicago.org", "data.seattle.gov"], "maxRows": 20 }

Output

• query: every row is a raw dataset record with the dataset's own column names (dates as ISO strings, locations as objects). The run's key-value store gets a QUERY_SUMMARY record with the dataset name, licence, columns, the query that ran, rows pushed and nextOffset for continuing.
• datasetInfo and catalogSearch: one row per dataset:

{
  "type": "dataset",
  "domain": "data.cityofchicago.org",
  "datasetId": "ydr8-5enu",
  "name": "Building Permits",
  "description": "Permits issued by the Department of Buildings...",
  "category": "Buildings",
  "attribution": "City of Chicago",
  "license": "See Terms of Use",
  "rowsUpdatedAt": "2026-10-18T10:02:11.000Z",
  "rowCount": 865123,
  "columns": [{ "fieldName": "permit_", "name": "PERMIT#", "type": "text", "description": "Permit number" }],
  "url": "https://data.cityofchicago.org/d/ydr8-5enu",
  "apiEndpoint": "https://data.cityofchicago.org/resource/ydr8-5enu.json"
}

Pricing

Event	Price	Notes
100 rows returned (rows-100)	$0.01	Primary event. 10,000 rows = $1.00; a partial last block counts as one
Dataset described (dataset)	$0.002	datasetInfo and each catalogSearch hit

Plus Apify's standard start fee ($0.00005 per run). No proxies, no browser. A failing query charges nothing. Set a maximum cost per run and the Actor stops cleanly at the limit with the rows so far saved.

How to use it

• Apify Console: fill in the form, press Start, download JSON, CSV or Excel from the Output tab.
• API: POST https://api.apify.com/v2/acts/lathewright~socrata-open-data-query/run-sync-get-dataset-items?token=<token> with the input JSON as the body.
• n8n, Make, Zapier: use the Apify node/module, pick this Actor, paste the input JSON; use nextOffset from QUERY_SUMMARY to continue large pulls.

Use from an AI agent (MCP)

Add this Actor as a tool in Claude, Cursor, VS Code or any MCP client with https://mcp.apify.com/?tools=lathewright/socrata-open-data-query. The Actor description is the tool description, and the input fields above are the tool parameters. Example prompts: "Find Chicago datasets about building permits", "How many noise complaints did each NYC borough log this week?".

Data sources, robots.txt and fair use

• Portals are queried through the SODA API (https://<portal>/resource/<id>.json) and described through https://<portal>/api/views/<id>.json; the catalog through api.us.socrata.com/api/catalog/v1 (no robots.txt, public discovery API).
• Before touching a portal the Actor fetches its robots.txt, refuses any portal that disallows the API paths and honours Crawl-delay for all user agents (the Socrata portals checked on 19 October 2026, including data.cityofnewyork.us, data.cityofchicago.org, data.seattle.gov, data.cdc.gov and data.ny.gov, allow /resource/ and /api/views/ and publish Crawl-delay: 1, which the Actor applies as a minimum spacing of one second; a portal asking for a longer delay, such as 60 seconds, is honoured as well).
• Each portal's own terms and dataset licence apply to what you do with the rows; the licence and attribution fields are returned with every dataset description. Public datasets occasionally contain names of individuals (e.g. licence holders); you are responsible for how you use them.
• One request at a time per portal, 100 to 10,000 rows per request, your optional app token raises Socrata's per-IP throttle.

FAQ

Which portals work? Any Socrata-hosted portal (the URL has /resource/ and dataset ids like abcd-1234). ArcGIS Hub, CKAN and OpenDataSoft portals are different products and are not covered.

Why is my query rejected? SoQL errors (unknown column, bad quote) come back verbatim in the failure message. Use datasetInfo to see exact field names.

Can I download a whole dataset? Up to 100,000 rows per run; continue with offset = nextOffset for more. For multi-million-row bulk exports use the portal's export links instead.