CA Police Incidents - SF, Oakland & LA (anonymous)

San Francisco, Oakland and Los Angeles each publish their police incident feed as open data (city Socrata, free public data). This actor turns them into a queryable, charged-per-record API - for underwriting, claims, neighborhood-safety and due-diligence analysis.

Important: all three feeds are anonymous incident records - no personal names. This is a geographic / statistical crime layer, not an individual-screening tool.

What it covers

San Francisco (city: sf) - SFPD Incident Reports, 2018-present, rolling (~1.06M records): crime category / subcategory / description, resolution (arrest / open / closed), police district, analysis neighborhood, date, lat/lon.

Oakland (city: oakland) - CrimeWatch, 2008-present, rolling (~1.28M records): crimetype (e.g. BURG - AUTO, STOLEN VEHICLE, PETTY THEFT, DOMESTIC VIOLENCE), description, police beat, address. No coordinates.

Los Angeles (city: la) - LAPD NIBRS Offenses, 2024-present, rolling live (~466k records): the richest feed - NIBRS crime code + description, crime-against (property/person/society), premise, weapon, disposition, LAPD area, lat/lon, and flags for victim shot / domestic violence / hate crime / gang-related / transit-related.

Typical questions

• "What crimes happened in the Mission (SF) last month?"
• "Car break-ins / stolen vehicles in Oakland by beat - recent trend?"
• "Domestic-violence or hate-crime incidents in Southeast LA this year?"
• "Which categories dominate this city right now?"

Low cost

From $0.001 per record - billed only for the rows you use, with a per-run charge cap so a broad query can't surprise-bill.

Three cities, three very different datasets, one schema. SF's feed is rich (coordinates, ~1.06M incidents); Oakland's is thin (no coordinates, ~1.28M rows back to 2008); LA publishes under the national NIBRS standard with a different column set - and its table carries rows dated back to the year 1800, which we exclude automatically on every pull. Category matching, coordinate handling and exact-vs-substring semantics are handled per city, then normalized onto one query surface so a date-windowed or area-filtered scan reads identically across all three. You query one clean incident surface across SF, Oakland or LA; the three-city schema archaeology stays on our side.

Inputs

Input	What it does
city	sf (default) / oakland / la
category	SF category / Oakland crimetype / LA NIBRS description
dateFrom / dateTo	incident date window
area	SF neighborhood / Oakland beat / LA area name
district	SF police district
keyword	incident description substring
domesticViolence / hateCrime / gangRelated / victimShot	LA-only flag filters (yes/no)
aggregate / groupBy	one row per group (category / area / district / year) with a count
maxResults	cap records (default 200)

Default run = the ~200 most recent SF incidents - a live "what's happening now in SF" feed.

Source

• SFPD Incident Reports - official, daily refresh.
• Oakland CrimeWatch Data - official, rolling.
• LAPD NIBRS Offenses - official, rolling.