Portia AI —
Building multi-agent, stateful and authenticated workflows

We've been watching Portia AI since they emerged from stealth earlier this year, and frankly, we're impressed by how they've tackled one of AI's biggest headaches: making agents actually work in the real world. Most companies are still figuring out how to build flashy demos, but Portia went straight for the jugular with enterprise-grade reliability.

The Smart Approach

What caught our attention first was their obsession with predictability. Instead of throwing another black-box agent into the market, Portia built their framework around a simple but clever idea: separate planning from execution. This means you can actually see what an agent is thinking before it acts, and more importantly, you can step in and course-correct when needed.

The authentication layer is particularly smart. While everyone else treats AI agents like fancy chatbots, Portia designed theirs to work in regulated environments where every action needs a paper trail. That's the kind of forward-thinking that makes us think they actually understand enterprise needs.

Why We're Bullish

The founding team brings serious credibility from Amazon, Google, and Stripe. That's not just name-dropping, it shows they've dealt with production systems at scale before. When you're promising enterprise-grade AI infrastructure, that experience matters more than flashy research papers.

Their open-source approach is refreshing too. Instead of locking everything behind proprietary walls, they're building in public and letting developers kick the tires. The cloud platform complements the SDK nicely without feeling like a cash grab.

Who Should Pay Attention

If you're building AI agents for anything involving compliance, customer service, or financial workflows, Portia deserves a serious look. Their KYC and refund processing examples aren't just demos, they're exactly the kind of boring-but-critical use cases where reliable agents could save companies millions.

The $5-6M seed round from General Catalyst and others gives them runway to prove their thesis. We're curious to see how quickly they can turn their technical advantages into market traction, but the early signs are promising.