Authorization and evidence for AI agents that spend money. The rail asks before money moves. The control isn’t with the agent. Open source, Apache 2.0.
Brooklyn, NY
Joined October 2026
- Tweets2
- Following10
- Followers4
- Likes0
Pinned Tweet
Spend authorization for AI agents. The rail asks before money moves. A stolen agent key can’t spend on its own.
Open source. Proof, threat model and every bug found so far, in the thread 👇
Vibe-coded @getwaysafe an authorization layer for AI agents that spend money. Then had two adversarial reviews try to break it.
18 security bugs. Fixed every one, tested first.
3 attempts to spend with a stolen agent key. All rejected on-chain.
waysafe.ai/proof
Waysafe retweeted
Vibe-coded @getwaysafe an authorization layer for AI agents that spend money. Then had two adversarial reviews try to break it.
18 security bugs. Fixed every one, tested first.
3 attempts to spend with a stolen agent key. All rejected on-chain.
waysafe.ai/proof