I’ve always wanted to be able to tell an agent what bet to make let it handle the rest But I’d never trust it enough to run in dangerous-skip all permissions mode with real money on the line. And if I’m forced to manually approve every trade, what's the point? So we built this demo to show a mode of dangerous-skip permissions mode with no actual danger Prompt the bet you want to make, the agent finds what it believes is the right trade, and delta Mandate verifies the proposed trade against the instructions to evaluate block or release

Today we’re announcing @deltadotnetwork integration with @Google's AP2. Link in the comments below, but here's why we're excited: AP2 creates a verifiable chain for user authorization, merchant commitment, and payment intent. But it has two significant shortcomings: 1) it does not verify that the agent’s action satisfies the substance of what the user authorized 2) it requires issuers and merchants to blindly trust the platform's attestation without having any way to verify the user's purchasing intent AP2 with delta Mandate addresses both of these problems. The user authorizes an intent mandate, the agent discovers a product, and the merchant only signs the checkout after delta has verified the proposed purchase against evidence from the actual product. This check sits outside the agent, so issuers and merchants do not need to blindly trust the platform. This matters for liability. Issuer banks are not going to underwrite delegated purchases if they cannot independently verify that the agent stayed inside the user’s mandate. They need more than an audit trail showing what happened after the fact. They need a pre-release control that ties authorization, execution, and evidence together before money moves. AP2 gives agentic commerce a much-needed authorization and transaction artifact chain. delta Mandate adds the enforcement + evidence layer required for issuers, merchants, and wallets to trust that chain at scale.

Full announcement here:

Agents can now discover products and check out. But should you actually trust them with real money to buy the right thing? Our research shows that you shouldn’t, at least without delta Mandate. Without delta’s intent enforcement, agent’s bought the wrong item 28.8% of the time on average, and the failure rate increased to 71.4% failure rate for harder intents. With delta Mandate in place, wrong purchase rate fell to zero. This is the enforcement layer behind the Mandate Guarantee that we offer via the platforms we partner with — if something slips through our policy engine, we’ll make the user whole. Use the docs in the comment below to connect download our MCP endpoint and try it with your own agent. If you want to offer the Mandate Guarantee to your customers, reach out and we’ll get you onboarded!