Alchemy's AgentPay Wants to Be the Rosetta Stone for Squabbling AI Payment Systems

Alchemy has dropped AgentPay, a tool that aims to make the AI payment systems from Coinbase, Stripe, Visa, Mastercard, and Circle stop acting like divorced parents who refuse to be in the same room together.

Here's the thing: agentic payment systems are popping off faster than rug pulls at a conference afterparty, but they don't talk to each other. Not even a little. A merchant who wants AI agents as customers is looking at building a separate integration for every single protocol—like assembling IKEA furniture, but the instructions are in different languages and the tools are missing.

"That's not sustainable, and it's only going to get more fragmented as more systems launch," Alchemy CTO Guillaume Poncin said in an email. "AgentPay fixes that. A merchant registers their existing API with us, we give them a new endpoint, and any agent on any supported protocol can pay them through it."

AgentPay is promising one integration to rule them all, supporting x402, MPP, A2P, and L402. It's basically playing diplomat in a room full of payment systems that can't stop beefing.

"We sit in the middle as the translation layer, where AgentPay routes instructions, and Alchemy never touches the funds," Poncin said. So basically, they're the wingman who sets up the date but leaves before things get awkward.

The company is widely regarded as the "AWS of Web3," providing the infrastructure, developer tools, and node services that keep blockchain apps from crumbling like a house of cards in a wind tunnel. Agentic finance—poised to become the backbone of internet payments—involves micro and nano-transactions, often between AI agents with humans occasionally lurking in the background like that one friend who just watches everyone else play video games.

Alchemy has soft-launched AgentPay via private beta and is targeting a general release in the coming weeks. The degen era of AI-to-AI commerce is approaching, and apparently, someone finally realized we needed a universal translator before it all becomes an even bigger mess.