I used instinct to buy baby gifts for two friends. And shit happened.
One couple is having a boy, the other a girl. The agent asked if I wanted colors to lean boy/girl or stay neutral. I said “neutral is fine.” Both orders were paid and went through.
Then nothing shipped.
Guess what, each item needed a color. The agent never picked one, the merchant accepted the order anyway, and the fulfillment system couldn’t process it.
You see, @jbahrdestefano it’s not only you lol
Most of the companies out there built their flows for humans. Agents act differently, they guess, skip, or improvise. When that happens, shit happens.
If your company is accepting AI agents you must track all of their actions. Understand if those actions match against the intent of the user, preventing incorrect actions before they happen.
Easy to think about this use case by itself. But when you imagine thousands of agents every hour coming to your platform, dealing with this scale will be a challenge.
There are some cool folks working on this, if your company is facing any issues, talk to them. They're called @beltichq
@t_blom Everyone's building the agent. Almost no one is building the other side of the conversation.
That's the most underrated infrastructure gap right now. Businesses are about to get hit with millions of agents and they still can't tell which agents to trust.
That's what we're building at Beltic.
One week out of stealth. Biggest signal: companies don't want to block agents, they want to accept the right ones, on their terms, now.
That's what we're building at Beltic. Short guide on how we think about Know Your Agent: https://t.co/fwoM87dnsW
For too long, agentic identity has been bottlenecking the growth in agentic commerce, as merchants didn't have a way of knowing which agents to trust, and whether they were authorized to enter certain transactions.
We're thrilled to invest in the Seed round and support @Isha@mikeallanpe and the incredible team at @beltichq as they solve this critical problem.
"How do I know this is really my customer's agent? How do I know it's allowed to take this action? How do I stop the bad ones without blocking the good ones?"
That's Beltic co-founder @mikeallanpe, and it's the entire agent-commerce trust problem in three questions.
@NorwestVP just led a $7.3M seed into the company built to answer them.
@beltichq, out of San Francisco, is agent verification infrastructure. Their KYA protocols — know-your-agent — verify an agent's identity, its authority, and the transactional legitimacy of what it's doing, in real time. The idea: businesses, banks, and platforms can authorize autonomous transactions and stop fraud before it clears, no matter what protocol or payment rail the agent runs on. $8.8M raised in total.
The founding trio reads like a hiring spec for this exact problem. CEO @Isha spent six years at Coinbase building the identity and onboarding infrastructure behind 50M+ users and its first institutional clients. @mikeallanpe founded Atar, a banking-as-a-service provider across Brazil's payments and core banking systems, acquired by Porto Seguro. CTO @farhanafsahi co-founded Verifiet — which Beltic itself acquired in 2025 — a global entity data infrastructure spanning 500M+ entities.
They're already working with a select group of design partners at the front of agentic commerce and payments, giving early access to the infrastructure and helping shape the standards for how agents transact.
What caught my eye: eight days ago, Baselayer raised $35M to extend its bank-identity network into "Know Your Agent."
Now Norwest seeds the company that IS the KYA layer — native, rail-agnostic, built for agents from day one. Two bets on the same layer in eight days. The market seems to be pricing the identity of the transacting agent as its own layer.
Congrats Isha, Mike, Farhan, and the entire Beltic team.
Up and to the right my friends! 💥🚀💪
🤖 Beltic is out of stealth with $7.3M seed led by Norwest Venture Partners, with Restive Ventures, Oxford Seed Fund, and Collide Capital.
The San Francisco startup verifies which AI agents hit your site and what they're allowed to do — "know your agent".
Excited to share that @NorwestVP is leading @beltichq's seed round as the company emerges from stealth today.
As AI agents begin transacting on behalf of businesses and consumers, financial institutions and merchants will need to know more than who an agent is. They’ll need to understand what it’s authorized to do, on whose behalf, and who is accountable. We believe that creates an entirely new infrastructure layer for agentic commerce.
That’s what Beltic is building.
@Isha and @mikeallanpe have both lived this problem firsthand, with deep experience building identity, verification and financial infrastructure at scale.
Beltic sits at the intersection of two areas where we have strong conviction: the modernization of financial infrastructure and the emergence of agentic commerce.
We’re thrilled to partner with Isha, Mike and the rest of the Beltic team on this important mission!
An AI agent can buy something in seconds.
Can the business receiving that order tell who sent it, what it can do, or whether to let it through?
We built @beltichq to answer that. We're out of stealth with $8.8M raised to make agent traffic verifiable.