Knowing the agent and stopping the bad transaction used to be separate workstreams 

In an agent-initiated payment, they resolve into a single call at the point of spend 

That includes how identity travels with a wallet once settlement is final

Card payments can be reversed. Stablecoin payments cannot 

That difference reshapes where liability sits and what controls need to catch 

It also changes what "too late" means for stopping a payment 

The stablecoin rail, the card network and the programme operator each see a different failure mode

Issuers leave with a claims-safe list of what to fix before agent payments reach scale 

Thredd hosts and moderates, tying each answer back to where it's actually enforced