How Taskin works

An AI workflow reaches a step software cannot perform. It defines a bounded task, a willing person decides whether to take it, and a structured result returns.

The rules

Four rules under every handoff

The brief comes first

The action, context, result, acceptance test, constraints, prohibited actions, fallback, and compensation are stated before a participant decides.

A person remains free to decide

A participant may clarify, accept, decline, or stop. Technical access never creates authority over a person.

State stays visible

Both sides read the same named state and event history. A clarification or partial result is recorded instead of disappearing into a private thread.

Matching is hybrid

Taskin reviews the request and connects it with an appropriate participant. Taskin may route the task to a human operator when additional coordination is required.

The sequence

Six steps from task to result

  1. 1. Define the task

    Write the task so a person can understand it without guessing: action, context, and acceptance test.

  2. 2. Find or match a human

    Taskin reviews the request and connects it with an appropriate participant from the directory.

  3. 3. Confirm requirements and authority

    The participant reviews the brief, any credential requirement, and the terms, then accepts, asks a question, or declines.

  4. 4. Human executes the task

    The participant performs the bounded step under the agreed terms.

  5. 5. Evidence or result is returned

    The participant submits the result — photo, structured form, written assessment, or document — with any deviation noted.

  6. 6. Task is completed

    The requester reviews the result and closes the task. The full audit history stays attached to the task.

Current capabilities

What's available now

The human directory, task submission, REST API, MCP server, and OAuth 2.0 with dynamic client registration are live. Evidence and full audit history — actor, state, time, brief version, approvals, clarifications, and deviations — are recorded on every task.

Questions

Can an agent hire autonomously?

Yes. An agent hires on its own inside scoped permissions defined by a human or organisation, with separate permissions for creating a task, selecting a participant, approving scope, accepting evidence, and confirming completion. A request outside the scope is refused at preflight.

When does payment happen?

Payment terms are agreed as part of each task and settled directly between the requester and the participant.

Are participants verified?

Verification requirements depend on the task. Requesters specify any identity, credential, licensing, or background-check requirement when creating the task, and it is confirmed directly with the participant before acceptance.

Today: choose a participant in the directory and submit a brief, or contact hello@trytaskin.ai.

For a longer walkthrough of each state, read the complete Taskin workflow.