A new startup, Lava Payments, aims to take on payment giants by building a solution for the modern web where AI agents now handle transactions for their customers. The idea came to founder Mitchell Jones after he left his earlier Y Combinator-backed fintech startup, Lendtable, as he began to experiment with AI.
He saw the potential to build out a system that would make using AI and agent payments simpler and more developer-friendly. While experimenting with an AI app and trying to build what he thought was something simple, he realized he quickly spent more than $400 trying to build a basic form-filling agent.
“I kept running into the same issue,” he told TechCrunch. “I was using the same underlying models and tools again and again, but through different wrappers or platforms.” And each time, he had to start a new subscription, re-authenticate, and pay separately, “even though I was already paying for access to the core model.”
“That felt fundamentally broken,” he continued. “I didn’t want to keep rebuying access to the same thing under a different wrapper. What I wanted was a single wallet, one set of credits, and the ability to move between tools and providers without starting over every time so I could pay for what I was using.”
He decided to launch Lava Payments as a solution.
Lava is a digital wallet that lets merchants use usage credits to facilitate transactions.
The idea is that one set of credits working across merchants and services makes it easier for autonomous agents to make payments without needing human intervention. It works like this: A merchant can enable the Lava wallet for their customers to use and upload (credits) money to. Once a customer does that, they can take that money and use it at any merchant that also accepts Lava and any of the foundational models, like GPT and Claude, on a “pay as you go basis,” Jones said.
So, rather than having to pay for each tool, a user buys a one-time usage credit that AI agents can simply charge as they perform various tasks. No more asking the user to approve transaction after transaction.
“Without Lava, agents can’t move smoothly through the internet because they constantly get blocked when it comes time to pay,” he said. He used Google as an example, saying every time a person opens Google Maps, they don’t have to pay Google for that map, as they’ve already paid Verizon and AT&T to access the internet.
On Wednesday, the startup announced a $5.8 million seed round led by Lerer Hippeau.
Others in this space include startups like Metronome.
“We see the world as very interconnected,” Jones said about what makes his product different. “What we’re really focused on is building [for the] agent-native economy.”
Born to a working family in Dayton, Ohio, Jones said his parents always told him the best way to get ahead was to work hard, save money, and get a good education.
“You know, a lot of the things that most people are told,” he recalled, when speaking with TechCrunch.
Jones took that advice to heart. He got a good education (Yale), held some good jobs (Goldman, Meta), and then founded some companies (the fintechs Parable and Lendtable, the latter of which was YC S20).
Jones said he met his lead investors for Lava because he went to high school with Will McKelvey, now an investor at Lerer Hippeau. He said McKelvey has been following his career for a while and always wanted to work together someday, and Lava Payments was that someday.
Others in the round included Harlem Capital, Streamlined Ventures, and Westbound. The fresh capital will be used for hiring, building products, and developing go-to-market strategies.
Overall, Jones is ready for Lava to be the “invisible layer that kind of powers the AI web,” he says, especially as AI agents find themselves more and more in the checkout line.
“We should be enabling agents to move, transact, and build without friction,” he said.
“We want to make sure that AI is something that can be used by every single person, even a kid from Dayton, like myself.”
The title of this piece was updated to properly reflect what the company does.
