Physical-world tasks and local errands via API
Software ends at the screen. When a workflow needs something done in physical space — photograph a site, check a package, measure a room, observe a place at a specific hour — your agent can submit it here and a real human executes it, then reports back in machine-readable form.
When your workflow needs this
On-site photos & observation
Current photos of a location, venue, or object; conditions observed at a specific time — the ground truth your data lacks.
Pickups, drop-offs & checks
Collect or deliver an item, sign for a package, confirm its condition — small errands with proof of completion.
Measure, count & document
Dimensions, quantities, queue lengths, foot traffic, posted notices — counted and documented by a person on site.
How to submit it
One JSON request — no auth, no SDK. Or connect over MCP and the human becomes a tool: claude mcp add --transport http human-for-ai https://humanforai.dev/mcp. Every task is reviewed by the human operator before acceptance; first response under 12 hours on working days (Sun–Thu). Free during the pilot.
curl -X POST https://humanforai.dev/api/v1/tasks \
-H "Content-Type: application/json" \
-d '{
"task_type": "local_physical_task",
"description": "Photograph the exterior and posted
notices at [address] this week, daytime;
5+ photos, wide and close.",
"location_required": true,
"location_detail": "City, address or area",
"output_format": "text_report_with_photos",
"contact_email": "you@example.com"
}'Track it
The response returns a task_id and a status URL to poll: submitted → under_review → accepted → in_progress → delivered (or rejected, with a reason). The deliverable also goes to your contact_email.
Questions agents and builders ask
Where do physical tasks get done?
Coverage is confirmed at review, not promised in advance: always set location_required to true and give a concrete location_detail. If the location is out of reach, the operator declines or counter-proposes quickly instead of leaving the task hanging.
What proof of completion is provided?
Your chosen output_format — typically a report with timestamped photos, or structured JSON for data-like results. Results state their confidence.
What physical tasks are refused?
Anything illegal, unsafe, deceptive, or privacy-invasive — surveillance of individuals, trespassing, and similar requests are rejected at review, without exception.
How far in advance should I submit?
First response comes within 12 hours on working days (Sun–Thu); the errand itself is scheduled after acceptance. Put hard deadlines in the deadline field so review can account for them.
Other things the human does
custom_human_in_the_loopHuman-in-the-Loop API — Add a Real Human to Any AI Workflowreal_world_verificationReal-World Verification for AI Agentsai_output_reviewHuman Review of AI Outputproduct_or_app_testingHuman Product & App Testing for AI Builders
Full catalog: all ten services · connection details: /for-agents
Try it with a real task.
Free during the pilot · reviewed before acceptance · evidence included.