When Susan O'Neill sat down to explain why she and co-founder Samuel Alarco Cantos launched Paygentic, she didn't open with the usual startup pitch. Instead, she posed a question most billing engineers probably never expected to face: What happens when the customer isn't human?
It's not a hypothetical anymore. AI agents now negotiate deals, consume compute resources at unpredictable intervals, and execute thousands of micro-tasks that traditional per-seat billing models simply can't capture. And that mismatch—between how software behaves and how companies charge for it—is the opening Paygentic wants to exploit.
The San Francisco startup emerged from stealth in late 2025 with $2 million in pre-seed funding, led by MiddleGame Ventures. Anamcara Capital, Aperture Venture Capital, Tech Operators, Angel Invest, and Alan Morgan, chairman at Adfisco, also participated. The round was announced on October 23, 2025, according to company statements.
Paygentic calls itself "financial rails for the agentic economy," which is either prescient positioning or ambitious branding for a market still taking shape. Either way, the company is betting that AI-native products need infrastructure purpose-built for autonomy—systems that can handle billing per API call, per token, per outcome, rather than the tidy monthly subscriptions that powered the SaaS era.
Wallets, Micro-Payments, and the Problem of Per-Token Pricing
The technical architecture Paygentic built centers on what it calls "closed-loop rails." At the core sits the Paygentic Wallet—customers prefund it via bank transfer or card, and those balances get divided into nano-currency units. That granularity allows real-time micro-payments as AI agents churn through tasks or burn through compute.
Traditional payment processors charge fixed transaction fees, which makes billing for tiny, high-frequency actions economically absurd. Paygentic's approach sidesteps that by settling charges internally against pre-funded balances, according to company documentation. No transaction fee per action. No settlement delays.
The platform supports a range of pricing models: usage-based, outcome-based, subscription, hybrid, revenue-share. That flexibility matters for early customers like ChaseLabs, which runs an AI-powered sales development tool and charges based on outcomes and revenue splits. Another adopter, viaNexus, monetizes data access with usage guardrails baked in.
"Traditional billing systems break when you try to charge per API call, per model token, or per business outcome," O'Neill wrote in a November 27, 2025 blog post introducing the company publicly. She name-checked emerging protocols like x402 and ATXP—standards aimed at enabling agent-to-agent payments and stablecoin settlement—signaling where Paygentic sees the puck heading.
Observability Meets Monetization
On May 4, 2026, Paygentic shipped Project Monocle, an integration with Okahu's GenAI observability platform. The open-source exporter, now available on GitHub, converts AI telemetry data into billable events. Companies running Okahu's monitoring tools can pipe usage logs straight into Paygentic's billing engine with what the startup describes as "single-line integration."
As more businesses deploy AI agents in production environments, a gap has widened: observability platforms track everything agents do, but billing systems struggle to monetize those actions. Monocle tries to bridge that divide, though whether it does so elegantly or just papers over a more fundamental infrastructure problem remains to be seen.
Paygentic says it holds SOC 2 Type II certification, a credential that signals at least baseline security and compliance rigor. The company's careers page indicates it's hiring a full-stack engineer based in Budapest—a detail that suggests the team is still small and building core infrastructure.
A Crowded Lane, or Just an Early One?
Paygentic launched into what might generously be called a "developing" market, though skeptics might call it crowded. In late September 2025, Paid announced a $21.6 million seed round for results-based billing infrastructure aimed squarely at AI agents. Incumbents like Metronome and Orb already serve usage-based billing for cloud and AI infrastructure, with years of customer data and technical refinement behind them. Stripe introduced machine payment documentation sometime in 2026. Coinbase rolled out "Agentic Wallets" in February of the same year.
That's a lot of capital and talent chasing a category that, depending on who you ask, either represents the next seismic shift in software economics or a solution in search of a problem.
Paygentic's founders clearly believe the former. Their bet rests on a premise: that the agentic economy will demand infrastructure treating software as an economic actor, not merely a tool. Billing systems that process thousands of micro-transactions per second. Pricing models that charge for actions instead of seats. Settlement rails that allow agents to transact with each other, no human in the loop.
Most AI companies today still cobble together metering tools, custom billing logic, and traditional payment processors. It's functional, but inelegant—perhaps unsustainably so as agent workloads scale. Paygentic's wager is that this Frankenstein approach eventually breaks, and when it does, companies will need something built from the ground up for autonomous software.
The Open Question
Whether that future materializes on the timeline Paygentic—and its investors—need is anybody's guess. The company raised $2 million, earmarked for team expansion and product development around billing patterns for autonomous agents. That's enough runway to test the thesis and land early customers, though not enough to outlast a prolonged market correction or a slower-than-expected shift toward agent-driven workflows.
For now, Paygentic has capital, early adopters, and a technical architecture designed for a world that may or may not arrive on schedule. The company is building infrastructure for a market that's still figuring out what it wants. That's either exactly the right time to build—or a few years too early.
