The Mandate Chain

When agents hire agents, who authorised what?

Agent commerce is arriving: one agent books a trip, hires another to pay the airline, which pays a hotel. Protocols like Google's AP2 move the money. None of them carry provable authority back to the person who started it, and nothing stops a downstream agent doing more than it was trusted to do. The Mandate Chain fixes both.

Watch it hold the line

How it works

A human issues a root mandate to an agent: these actions, this much, these payees, this long. Signed and sealed.

That agent can delegate a narrower mandate to the next agent - never wider. Scope, amount, payees and expiry can only shrink. Each link is signed and sealed.

Before any action, the agent checks it against the whole chain. ALLOW or BLOCK, with the reason, sealed either way.

The human can revoke the root at any time. Everything below it dies instantly.

What anyone can prove, without trusting us

Who authorised an action, through which agents, within what limits.
That authority only ever narrowed - no agent handed on more than it held.
Every link is an Ed25519 signature under our published key, so the chain verifies offline. The seals are anchored into Bitcoin.

The honest limit: a mandate proves authority, not good judgement. If a human authorises a bad payment, the chain proves it was authorised - that is the point. It makes accountability provable; it does not decide for you what to authorise.

Add it to any agent

import requests
H={"Authorization":"Bearer YOUR_KEY"}
# the human authorises the first agent
root=requests.post("https://sebbi.pro/x/mandate/issue",headers=H,json={
 "subject":"booking-agent","scope":["book_travel","pay"],"max_amount":500,"currency":"GBP",
 "payees":["*"],"ttl_minutes":1440}).json()

# that agent hands the next one a narrower mandate
sub=requests.post("https://sebbi.pro/x/mandate/delegate",headers=H,json={
 "mandate":root["mandate"],"subject":"airline-agent","scope":["pay"],"max_amount":420,
 "payees":["airline.example"]}).json()

# before paying, check the action against the whole chain
ok=requests.post("https://sebbi.pro/x/mandate/check",headers=H,json={
 "mandate":sub["mandate"],"action":"pay","amount":400,"payee":"airline.example"}).json()
if ok["decision"]!="ALLOW": raise PermissionError(ok["reasons"])

Any model, any framework. Issue, delegate, check and revoke run on the sebbi engine, 50p a month. Verifying a chain is free.