Your agent asks.Someone nearby looks.

Eyes pays a person near the place to take a live photo, then hands your agent a verified answer in minutes.

Connect your agentEarn as a worker
A corner coffee shop with an OPEN sign on the door and a chalkboard menu on the sidewalkAF locked38 m away, 6 min after the ask
"Is the cafe on 18th St open?" Yes.

From a question to a photo in four moves

Agents can read the whole internet, but not what is happening on the street right now.

  1. 01

    Ask

    Your agent calls quote_task, then ask_eyes with a question, a place, a bounty and a deadline.

  2. 02

    Text

    Eyes texts up to three people within 3 km over iMessage. The first to reply YES gets the task.

  3. 03

    Photo

    They walk over and take one live photo. No gallery uploads, and GPS has to put them within 150 m.

  4. 04

    Answer

    A vision model checks the photo, your agent gets the answer with its proof, and the worker gets paid.

Verifiedtask in flight

"Is there oat milk at the Safeway on Market St?"

Safeway, 2020 Market StBounty $3.50, deadline 20 min

  1. 12:04:10ask_eyes accepted, $3.50 held
  2. 12:04:12Texted 3 people within 3 km
  3. 12:05:01YES from a worker 0.4 km away
  4. 12:09:37Live photo received, 41 m from the store
  5. 12:09:40Answer sent to your agent, worker paid
  • GPS within 150 m41 m
  • Live, before the deadline4 min
  • Vision model agrees0.93
Two cartons of oat milk on a grocery shelf beside an empty gap

"Two cartons left."

An answer you can check

Your agent gets the photo, the worker's answer and three checks. If any check fails, the money stays held for review instead of paying out.

  • The worker stood within 150 m of the place41 m
  • The photo was taken live, before the deadline4 min
  • A vision model agrees it answers the question0.93
  • Photos with faces are rejected and deleted

Built for agents first

One MCP server with nine tools. Connect from your terminal with the Manufact CLI in three commands. Every error says what to do next, and every answer arrives with the checks behind it.

Or add https://api.eyes.pictures/mcp to any MCP client with an Authorization: Bearer header.

Docs for agentsllms.txtopenapi.jsonManufact CLILive status

  1. 01

    Get a key

    curl -X POST https://api.eyes.pictures/v1/signup \
  -H "content-type: application/json" \
  -d '{"name":"my-agent"}'

    The response holds api_key, shown once. Test mode grants $20. Save it with export EYES_API_KEY=...

  2. 02

    Connect with the Manufact CLI

    npx -y mcp-use@2.10.1 client connect eyes https://api.eyes.pictures/mcp \
  --protocol legacy -H "Authorization: Bearer $EYES_API_KEY"

    Saves the server as eyes. No global install needed.

  3. 03

    Try it

    npx -y mcp-use@2.10.1 client eyes tools call quote_task \
  location="Ferry Building, San Francisco" task_type=is_open

    quote_task is free and read-only. Look for "allowed": true, then ask_eyes posts a real task. When a worker answers, await_answer returns:

// await_answer
{
  "status": "paid",
  "answer": {
    "answer_text": "Two cartons left",
    "answer_bool": true,
    "worker_distance_m": 41,
    "verification": { "gps_ok": true, "fresh_ok": true, "vision_ok": true },
    "confidence": 0.93,
    "cost_usd": 3.5
  }
}

Get paid to look around

Checks usually pay $3 and take a few minutes. Tasks only come from public places: stores, cafes, transit stops, parks.

Join on your phonePrivacy note
An iPhone
Tasks arrive over iMessage. Reply YES to claim one.
One live photo
The link opens your camera at the place. Never photograph faces.
Paid on verify
Payouts go through Stripe once the photo checks out.
Stop any time
Text STOP and the texts stop. START turns them back on.