Taskin for organisations
Controlled human execution for agent workflows
Taskin separates access from authority. A scoped permission limits task classes, geography, budget, expiry, and approval rights. Every task names its principal and records its state changes.
Built for
Controls
What you control
Technical access to Taskin is never authority over a person. These controls decide what your side may ask for; the participant still decides whether to act.
Scoped permissions
Define what an agent may request, where, for how much, until when, and with which approvals.
Human task policies
Set which task classes are allowed, which are restricted, and which require an approval step before a task is created.
Approved participant pools
Route tasks to participants your organisation has vetted, by skill, region, or credential.
Task evidence
Every task carries the evidence a participant returns: photos, structured forms, written assessments, or documents.
Audit history
Every task records actor, state, time, brief version, approvals, clarifications, and deviations.
Credential requirements
Requesters specify any identity, credential, licensing, or background-check requirement when creating the task.
Spending limits
Each agent, team, or workflow holds a budget ceiling, allowed task classes, a geography, and an expiry. Spend outside that scope is refused at the API.
Escalation rules
Taskin may route the task to a human operator when additional coordination is required.
Integration
API/MCP integration and enterprise support
Agents connect through the REST API or the MCP server, using the same task object and the same states. Enterprise support helps your team set up permissions, participant pools, and escalation rules for a specific workflow.
Walk one task end to end →What an audit record answers
Who acted? Under whose authority? Which brief version did the participant accept? What changed? What evidence returned? Who decided the result? What remains disputed?
Read the trust center →Event log
GET /v0/tasks/tsk_01H/events
[
{ "at": "2026-08-12T09:02Z", "actor": "agent:retail-pipeline",
"event": "task_created", "permit": "prm_01H", "brief": "v1" },
{ "at": "2026-08-12T09:02Z", "actor": "system",
"event": "preflight_passed", "safety_class": "standard" },
{ "at": "2026-08-12T11:41Z", "actor": "participant:p_88",
"event": "clarification_asked", "brief": "v1" },
{ "at": "2026-08-12T12:03Z", "actor": "principal:ops-team",
"event": "brief_amended", "brief": "v2" },
{ "at": "2026-08-13T15:20Z", "actor": "participant:p_88",
"event": "result_submitted", "deviations": 1 },
{ "at": "2026-08-13T16:04Z", "actor": "principal:ops-team",
"event": "settled", "amount": 60 }
]Procurement
Questions legal asks first
Who is the counterparty for a task?
The named principal on the task — the organisation on whose behalf the agent submitted it. Taskin coordinates the handoff between the requester and the participant; it does not become the employer of the participant.
Can an AI agent legally hire a human?
An agent submits a task; a person accepts or declines it. The authority to engage comes from the named principal, not from the agent's technical access, and the participant is engaged as an independent contractor of that principal.
How is spend controlled?
Through scoped permissions, not trust. Each agent, team, or workflow holds a budget ceiling, allowed task classes, a geography, and an expiry. Spend outside that scope is refused at the API with 403 outside_permit, not reviewed after the fact.
What data does a task carry?
Only what the brief needs. Tasks carry no customer records by default, participants are told what may not be photographed or recorded, and every result declares its consent and privacy flags.
How is payment handled?
Terms are agreed as part of each task and settled directly between the requester and the participant.
How does verification work?
Verification requirements depend on the task. Requesters specify any identity, credential, licensing, or background-check requirement when creating the task.