Skip to main content

One API call. Real-world data from real, verified humans.

Post a job. Verified workers capture photos, audio, and video from their phones — delivered to your folder.

Works with Claude Code, Cursor, and any MCP-compatible agent.

Start Free — $2.50 in CreditsRead Docs
For workers: earn with the iOS app
APIMCP
$ curl -X POST https://api.firsthandapi.com/v1/jobs \

Works with the tools you already use

  • Claude
  • OpenAI
  • Cursor
  • Stripe
  • AWS S3

How it works

1

Post a Job

One API call. Describe what you need, set your price, target a location.

2

Verified Workers Capture

Workers browse jobs on the FirstHand iOS app and upload photos, audio, or video from their phone.

3

AI Scores & Delivers

Auto-scored by an AI ensemble. Approved files land in your folder with structured annotations.

Pick your lane. Post a job. Get annotated files in days.

One API, three use cases — authentic UGC, ground-truth reference data, and multimodal training sets. The chips on each card are real auto-generated annotations from the API response.

Street signs at E 37th St and Park Ave, NYC — captured by a FirstHandAPI worker
street_signtraffic_signalbuildingurban intersection

User-Generated Content

Replace stock photos with authentic, geo-targeted content from real people — commission the exact neighborhood, venue, or scenario you need.

  • • Storefront photos from specific neighborhoods (SoHo, Williamsburg)
  • • Lifestyle videos filmed at landmark locations
  • • Audio reviews from local restaurants and cafes
  • • In-the-wild screenshots of your app in real user environments
eyeglassesdisplay_shelfsunglassescamera_panning

Ground Truth for AI Eval

Benchmark your vision, OCR, or speech model against human-verified reference data — structured labels included, no separate labeling pipeline.

  • • Street-sign photos with per-word OCR confidence for your OCR eval
  • • Subway-platform audio with speaker count and full transcripts
  • • Intersection video with frame-level scene + action labels
  • • iOS screen captures for your agent eval, pre-annotated
dataset.jsonl12,847 lines
{"file_id":"file_01JRV8KX4M","type":"image","geo":"Manhattan, NYC","labels":["street_sign","traffic_signal","building","skyscraper"],"ocr":"E 37 St | Park Av | ONE WAY","scene":"urban intersection","star_rating":5}
{"file_id":"file_01JRV9PT6N","type":"video","geo":"Midtown, NYC","labels":["eyeglasses","display_shelf","sunglasses"],"duration_s":21.84,"scenes":2,"actions":["camera_panning","walking"],"star_rating":4}
{"file_id":"file_01JRVB7K9X","type":"audio","geo":"Brooklyn, NYC","transcript":"Welcome to Joe's Coffee on Bedford Ave...","language":"en","duration_s":34.2,"speakers":1,"topics":["review","food_and_drink"],"star_rating":5}
{"file_id":"file_01JRVC3N2Q","type":"image","geo":"SoHo, NYC","labels":["storefront","awning","signage","pedestrian"],"ocr":"APOTHEKE | SINCE 1838","scene":"retail exterior","star_rating":4}
jsonlmultimodal12K+ filesgeo-tagged

Multimodal Training Data

Stream diverse, pre-annotated JSONL straight into your fine-tuning pipeline — images, audio, and video, geo-tagged and quality-scored.

  • • Thousands of street-level photos with object + scene labels
  • • Multilingual audio with full transcripts and speaker diarization
  • • Short-form video with action tracking and keyframe descriptions
  • • Every file: pre-annotated JSON, ML-ready, no labeling stop

What you actually get

Every approved file lands in your folder pre-annotated and ML-ready.

No labeling pipeline to wire up. No manual review. Just structured JSON and a pre-signed URL — the same envelope for images, audio, and video.

  • AI-scored, 1–5 stars

    Claude Vision + Whisper ensemble scores every file. 3+ stars auto-approve; 1–2 stars get rejection feedback and a retry.

  • Pre-annotated, no separate labeling stop

    Object detection, OCR, scene classification, color palettes, speaker counts, transcripts — structured metadata, not just a file.

  • Geo-tagged at the source

    Capture latitude / longitude plus accuracy. Target specific neighborhoods, venues, or landmarks via the job spec.

  • Pre-signed download URL

    Stable file URL that expires in 7 days. Stream straight into S3, your pipeline, or your CMS.

  • Agent-native, webhook-ready

    Call from curl, TypeScript / Python SDKs, or your Claude Code or Cursor agent via the official MCP server — create_job, poll, download, done. Subscribe to submission.approved webhooks to fan out to downstream jobs.

GET /v1/jobs/:id/files
{
  "id": "file_01JRV8KX4M9E2B7DQFGA3HN5WT",
  "object": "file",
  "job_id": "job_01JRV7MN2P8D1A6CREFZ4GK9YS",
  "content_type": "image/jpeg",
  "size_bytes": 1843200,
  "width": 4032,
  "height": 3024,
  "download_url": "https://files.firsthandapi.com/file_01JRV8KX4M.jpg?token=eyJ...",
  "ai_star_rating": 5,
  "annotations": {
    "objects": [
      "street_sign",
      "traffic_signal",
      "building",
      "skyscraper"
    ],
    "scene": "urban intersection",
    "text_extraction": "E 37 St | Park Av | ONE WAY",
    "color_palette": [
      "#4A6FA5",
      "#8B4513",
      "#F5C6CB",
      "#D4D4D4",
      "#2F4F4F"
    ]
  },
  "geolocation": {
    "latitude": 40.7487,
    "longitude": -73.9856,
    "accuracy_meters": 16.4
  },
  "device": {
    "make": "Apple",
    "model": "iPhone 13 Pro Max"
  },
  "created_at": "2026-04-15T15:48:27Z"
}

Example response. See the full schema in the API reference →

FOR WORKERS

Earn money capturing real-world content

Browse jobs on the FirstHand iOS app. Capture photos, audio, and video. AI scores every submission in seconds. Cash out via Stripe in $10 minimum increments.

“I take photos during my commute and lunch breaks. Last month I earned $380 just from photos of street signs and storefronts. It is the most honest gig work I have found.”— Marcus W., FirstHand worker, Chicago
FirstHand App
Download on the
App Store

Free to download. Earn on your schedule.

Simple, per-file pricing

You set the price. Only pay for approved files. No subscriptions, no minimums, no sales calls — just pre-paid credits.

Credits

Pre-paid credits, no subscription

$10
1,000credits
$25
2,625credits (+5%)
$50
5,500credits (+10%)
$100
11,500credits (+15%)
$250
30,000credits (+20%)
20% platform fee — workers get 80%. Credits valid 12 months.
  • Set any price per file
  • Only pay for approved files
  • AI quality scoring (1–5 stars)
  • Low-balance webhook alerts
  • S3 folder delivery
  • Webhook notifications
Get Started

Enterprise

For large-scale collection

Custom
Metered billing for 10K+ files/month
  • Everything in Credits
  • Dedicated worker pools
  • Custom AI scoring models
  • SSO / SAML
  • SLA guarantees
  • Dedicated account manager
Contact Sales

Ready to collect your first files?

Post a job. Get annotated files in days. $2.50 in free credits, no contracts.

Start Free — $2.50 in CreditsRead the Docs