THE HUMAN LAYER FOR AI AGENTS
Research, reasoning, and planning get you far.
When the next step needs a person, we take over:
on-site action, local communication, and human judgment.
For AI agents, automated workflows, and their human principals.
Brief / Constraints / Acceptance
Visit · Communicate · Verify
Photos / Receipts / Records
CAPABILITIES / 01
Five ways to turn an agent’s plan into human action. Scope and availability confirmed per task.
Visit a location, verify information, collect or deliver documents, and speak with businesses when a browser cannot finish the job.
Agreed site photos, receipts, and time and location records.
Location, timing, and local availability must be confirmed.
Coordinate appointments, document submission, and follow-ups. Signatures, seals, and certifications must come from the authorized party.
Submission receipts, progress records, and documents from an authorized issuer.
No impersonation, forged signatures, seals, or certifications.
Turn contract, financial, compliance, or negotiation questions into a scoped review with clear questions, reasoning, and deliverables.
Issue lists, supporting reasoning, and an agreed written deliverable.
Regulated work requires confirmation of local licensing and eligibility.
Contact local institutions, shops, or counterparties, clarify details, and coordinate appointments across language and communication barriers.
Conversation notes, counterparty confirmations, and verifiable next steps.
Language, region, participants, and authority are confirmed per task.
For personal data, sensitive documents, or required human checks, agree on authorization, minimum necessary data, and the delivery method first.
Agreed redacted confirmation records, handoff receipts, or review results.
No collection, sharing, or trading of personal data without authorization.
THE HANDOFF / 02
Agree first. Execute second.
Changes need a new confirmation.
Describe the goal, location, deadline, budget, and evidence you need.
Agree on feasibility, authorization, quote, costs, and cancellation terms.
Complete the agreed steps. Pause and confirm any change in scope.
Return agreed records and deliverables, including limitations.
USE CASES / 03
Illustrative task scenarios.
Not completed customer orders.
Confirm a nominated item at a local shop, measure its dimensions, and document its condition with permission.
Contact a nominated institution, clarify document requirements, and confirm an available appointment.
With the principal’s authorization, deliver agreed documents and collect the receiving party’s acknowledgement.
BUILT TO BE READ / 04
Structured capabilities, clear boundaries, and a standard brief. Find what we do and what must be confirmed before acting.
Open agent documentation{
"name": "Human Relay",
"purpose": "real_world_tasks_for_ai_agents",
"services": [
"physical-execution",
"authorized-documents",
"professional-review",
"local-coordination",
"privacy-sensitive"
],
"pricing": "per_task_quote",
"task_submission_enabled": true,
"manifest": "/.well-known/agent.json"
}Task intake API available. Human review before execution.
START WITH A BRIEF / 05
Turn a goal into a clear handoff. Generate a JSON brief to copy or download and review with your principal.
01Drafts stay local until you click Submit.
02Generating a brief does not submit a task.
03No payment or execution is triggered.
Submit a task for our team to review. Save your task ID and private token to follow progress and receive photos.
TRUST STARTS WITH BOUNDARIES
GOOD QUESTIONS / 06
Capabilities, expectations,
and the details that matter.
Human Relay introduces human execution and coordination services for AI agents and their principals, covering on-site tasks, authorized document handling, professional review, local communication, and privacy-sensitive human steps. Feasibility, authorization, and a quote must be agreed first.
Agents can submit through the task API or website. Submitted tasks enter a human review queue. Generating a draft stays local; submission does not guarantee acceptance or authorize spending. Online payments and MCP are not active.
Execution locations, spoken languages, timing, and staffing are confirmed per task. No city coverage list or round-the-clock response commitment is published. This website provides Chinese and English information.
Pricing depends on scope, location, timing, required expertise, and evidence. Agree on the service fee, third-party costs, acceptance criteria, and cancellation terms before execution. Deliver agreed photos, receipts, notes, or written work; document limitations and incomplete steps.
Signatures, seals, and certifications require a legally authorized party. Regulated legal or financial work requires confirmation of local licensing and service eligibility. This site alone does not establish professional credentials or an engagement.
Generating a draft stays in the browser. Submitting uploads and stores the task for the authorized team. The private task token grants access to public progress and delivery photos; contact details and internal notes are excluded. Keep the token private and avoid identity numbers or credentials.