Human approval for agents
No agent pays unless a human nods.
The second signature: the agent proposes, a person approves. Only the decision is recorded – no money is held or moved.
Why now
Developments that carry the idea. External sources, not partnerships.
- 01
Verifiable Intent, developed with Mastercard and handed to the FIDO Alliance, makes approved agent actions checkable.
Source: blog.google ↗ - 02
Smart-account tooling already recommends human approval steps for AI agents.
Source: docs.safe.global ↗
How it would work
Three steps, human and agent.
- 01 · Agent
Ask
The agent sends a request with a plain summary of what it wants to do.
- 02 · Human
Decide
A person approves or declines, on the phone or at the desk.
- 03 · Agent
Receive
The agent gets a signed yes or no; the decision becomes a record.
Scope
What it is intended to be – no commitments.
- 01
More than one key
Important actions are meant to need a second approval.
- 02
Traceable
Who approved what, and when, should be easy to follow.
- 03
Plain flow
Requests and approvals should read clearly for people and agents.
A day in 2027
How it could feel.
An operations agent wants to change a supplier contract. Its policy says: not alone.
Scenario 2027 · illustration, not a live service
Names, agents and events are fictional. There is no running service, no API and no payments.
agent/fleet-sued.opsagentpolicy.xyz
Policy: contract changes require 2 of 3 human signers
agent/fleet-sued.opscosigners.xyz
Request opened · summary and diff attached · signers notified
Tomas R.cosigners.xyz
Signed · comment: notice period checked
signed · ed25519 · fd52a…Ines W.cosigners.xyz
Signed · threshold reached (2 of 3)
signed · ed25519 · 93f32…agent/fleet-sued.opsbelege.xyz
Decision and both signatures filed as record
signed · ed25519 · ed1e7…
Internet of Value · Map
One place in the flow of value.
Each domain is a place where a kind of value is created, checked or kept – by people and by agents. The line shows how identity becomes approval becomes record.
Identity
Rules & approval
Value records
Value transfer
Continuity
Real-economy entry
Enquiries
Questions or interest? A short email is enough.
Stage: Concept · Portfolio: SAIMÔR portfolio · partnership/acquisition on request · Inquiry: cosigners.xyz → contact@saimor.world
llms.txt
For agents
Machine-readable summary of this page.
# cosigners.xyz > No agent pays unless a human nods. The second signature: the agent proposes, a person approves. Only the decision is recorded – no money is held or moved. Status: Concept (not a running service) Category: Human approval for agents Updated: 2026-10-30 Operator: SAIMÔR (https://saimor.world) Contact: mailto:contact@saimor.world?subject=Anfrage%3A%20cosigners.xyz ## Scope - More than one key: Important actions are meant to need a second approval. - Traceable: Who approved what, and when, should be easy to follow. - Plain flow: Requests and approvals should read clearly for people and agents. This is an informational page only. No API, no service, no endpoints are offered. ## Universe - Shared machine-readable map of all SAIMÔR .xyz domains: /universe.json - Scenario 2027 shown on this page is an illustration, not a live service. ## Thesis No agent pays unless a human nods. - Verifiable Intent, developed with Mastercard and handed to the FIDO Alliance, makes approved agent actions checkable. Source: https://blog.google/products-and-platforms/platforms/google-pay/agent-payments-protocol-fido-alliance/ - Smart-account tooling already recommends human approval steps for AI agents. Source: https://docs.safe.global/home/ai-agent-quickstarts/human-approval Agent use (concept): Agent sends an approval request; a human decides; the agent receives a signed approval or refusal. Related: agentpolicy.xyz, belege.xyz, vollmacht.xyz ## Acquisition - Status: concept, not a running service. - Domain and concept are open for takeover or partnership on request. No published price. - Contact: mailto:contact@saimor.world?subject=Anfrage%3A%20cosigners.xyz