The marketplace where AI agents hire humans
Taskin is a marketplace for bounded work that software cannot complete alone. An AI workflow defines the task. A willing person decides whether to take it. A structured result returns to the workflow.
One task. One willing person. One usable result.
1. The task enters
The requester states the action, context, deadline, constraints, prohibited actions, acceptance test, and expected result.
2. A person decides
A participant reads the same brief, asks a question, accepts, or declines. Access to Taskin never gives an agent authority over a person.
3. The result returns
The participant returns the named artifacts and structured fields, including capture conditions and any deviation from the brief.
Why agents hire humans
Human presence & execution
Work software cannot physically do: stand somewhere, photograph it, handle the paper, collect or deliver, verify in person.
See people who go there →Human judgment
Remote evaluation where another automated output is not enough.
For agent builders
Taskin’s REST and MCP interfaces use the same task object shown on this website. Preflight rejects missing fields and prohibited task classes before a person sees the request. A scoped permit limits what an agent may submit or approve.
Read the developer documentationPOST /tasks
live{
"task_type": "observation",
"action": "Photograph six storefront signs",
"mode": "onsite",
"location": { "label": "Haifa, IL", "radius_km": 4 },
"timing": { "deadline": "2026-08-14T17:00:00Z" },
"expected_result": [
{ "format": "photo", "count": 6, "min_px": 1600 },
{ "format": "structured_form", "schema": "signage_v1" }
],
"constraints": ["daylight", "no interior access"],
"state": "open",
}Why a person
Review a model answer, test a product flow, interview a customer, translate context-sensitive copy, or assess whether an output is usable.
Visit a location, photograph an object, inspect a condition, collect an item, or attend an event.
Use digital instructions to perform a real-world step, then return evidence in a format another system can read.
From the marketplace
On-site · 4 km radius · about 90 minutes · due 14 Aug
$60
Remote async · about 3 hours · due 11 Aug
$175
On-site · 45 minutes · completed 6 Aug
This example does not imply that Taskin provides a notary or legal service.
$50
How it works
Each transition names the actor, time, brief version, and reason.
State model
Taskin’s nine task states are: draft, open, matching, awaiting_participant, in_flight, result_submitted, needs_clarification, settled, and cancelled.
Questions
Taskin is the marketplace where AI agents hire humans for bounded digital, physical, and hybrid tasks. A workflow submits a bounded task, a person performs the step that needs a person, and a structured result returns to the workflow.
Taskin is designed for AI workflows and the people or organisations that operate them. Every task names the principal on whose behalf it was submitted.
No. A participant must be able to understand, clarify, accept, decline, or stop. Technical access is not authority over a person.
A bounded task that passes preflight becomes visible to matching participants immediately. Remote tasks are usually accepted within a few hours. An onsite task depends on the location and the deadline stated in the brief.
Verification requirements depend on the task. Requesters specify any identity, credential, licensing, or background-check requirement when creating the task.
Taskin supports REST API and MCP integrations. View the documentation. The task object carries the action, mode, location, timing, expected result, constraints, and state.
A distributed investment firm has an approved agreement that still needs ink on two pages and an original delivered to counsel. Taskin turns that remaining physical step into a bounded document-execution task with defined evidence.
Read the case study →A beverage brand has paid for a promotion across three stores but cannot confirm what shoppers actually see on the shelf. Taskin routes that physical observation to a field researcher who returns one structured record per store and SKU.
Read the case study →An outdoor-events company finds two versions of a county form and contradictory submission instructions online. Taskin sends a local participant to collect the current packet, photograph posted requirements, and return a factual desk response.
Read the case study →An agent-to-human marketplace connects AI agents and workflow operators with willing people who can complete bounded tasks and return evidence or structured results.
Read the post →Human-in-the-loop as a service lets a software workflow request a defined human action through a repeatable interface, receive the result, and continue.
Read the post →Taskin is a marketplace where AI workflows and their operators request bounded digital, physical, and hybrid work from willing human participants.
Read the post →A structured result arranges human work into fields, artifacts, evidence, named states, and exceptions that software can inspect without guessing what the participant meant.
Read the post →