Orthogonal is building the checkout counter for AI agents

Ask most founders what their company does and you get a noun - a platform, a tool, a network. Ask the team at Orthogonal and you get three verbs: discover, orchestrate, pay. Those verbs describe the small, unglamorous gap between an AI agent that can talk and an AI agent that can actually get something done. Booking a flight, enriching a lead list, scraping a page, running a search - each one means calling some outside service, and each service wants a signup, an API key, a credit card, and sometimes a sales call. A human can grind through that. An agent can't.

Orthogonal, a member of Y Combinator's Winter 2026 batch, exists to erase that friction. Through a single integration - an SDK, a REST API, a Model Context Protocol (MCP) server, or a command-line tool - an agent can reach 35+ paid APIs, and Orthogonal quietly handles the key management, the authentication, the routing, and the billing underneath. No subscriptions. No sales calls. The agent describes what it needs; Orthogonal finds the service, runs the call, and charges by the call, sometimes as little as a penny.

01 / The ProblemAPIs weren't built for agents

The last decade of software ran on a simple contract: a human signs up, remembers a password, enters a card, and clicks through a monthly subscription. That contract quietly assumes a person is in the loop. Strip the person out and it falls apart. An autonomous agent doesn't have a company credit card, doesn't want a yearly plan for a service it will call twice, and can't schedule a demo with a sales rep. It needs to find a capability, use it once, and pay for exactly that.

Orthogonal's co-founder Bera Sogut frames the missing pieces as three walls. Get past all three and you have something close to an autonomous agent. Leave any one standing and the agent stalls.

02 / The ProductOne wallet for the whole API economy

In practice Orthogonal behaves like a universal adapter. A developer wires an agent to it once. From then on, the agent can say, in effect, "I need company data on this domain" or "scrape this page," and Orthogonal's discovery layer surfaces the right provider from its catalog using natural-language search. Its orchestration runtime makes the call, handles the auth, and returns a normalized response. Its payment layer meters the usage and bills per call.

The early catalog leans toward the work agents do most: search, scraping, datasets, data enrichment, and identity verification. The providers already plugged in read like a who's-who of the data tooling world.

Payments are where the company's DNA shows. Orthogonal supports multiple rails: ordinary credits and fiat, x402 onchain micropayments, and MPP, an agent-native payment method. The point isn't crypto for its own sake - it's that a machine paying a fraction of a cent for a single API call is a different problem than a human paying $49 a month, and it needs rails designed for that.

03 / The FoundersPayments met APIs at McGill

The two founders met as students at McGill University in Montreal and stayed close, then spent years learning opposite halves of the same problem. Christian Pickett, the CEO, worked on payments at Coinbase and billing at Vercel - which is to say he learned that the hard part of payments is rarely the money itself, but the discovery, auth, and plumbing around it. Bera Sogut, the CTO, worked on reCAPTCHA and the Maps APIs at Google, and is a two-time ACM ICPC World Finalist. One knew how money moves. The other knew how APIs behave at scale. Orthogonal is what happened when those two halves finally shook hands.

The broader founding team pulls from Coinbase, Vercel, Google, and Amazon Robotics - a roster tilted heavily toward payments, APIs, and infrastructure rather than models. That's a tell about where they think the value sits. In a gold rush, they're building the checkout counter, not panning for gold.

04 / The MoneyA $4.3M bet on machine commerce

On June 25, 2026, Orthogonal announced a $4.3 million seed round led by Pantera Capital, with Y Combinator, Pioneer Fund, Decasonic, Blast, Outbound, and Surreal participating. The money is earmarked for building out the orchestration and payment layer so that agents can discover, invoke, and pay for services inside a single runtime.

05 / The MarketPicks and shovels for the agent economy

Orthogonal sits in a young, crowded corner of AI infrastructure. Adjacent efforts include Coinbase's x402 agent-payment work, and startups like Skyfire, Payman, and PayOS on the payments side, plus API aggregators such as RapidAPI and the growing pile of MCP tooling on the access side. What sets Orthogonal apart is that it refuses to pick just one of the three walls. Most rivals do payments, or discovery, or access. Orthogonal is trying to be the single runtime where all three happen at once - which is either overreach or exactly the integration point an agent needs.

Its customers today are the people building agents: AI engineers, go-to-market builders, and developers wiring autonomous workflows. Early users and testimonials include Nexio, Hobbes, and fellow YC W26 company MouseCat. The team is small - around six people - which is itself a bet that infrastructure this focused doesn't need an army, just the right primitive.

Whether the "more agents than people" future arrives on schedule is unknowable. But if even a fraction of it does, someone has to answer the plain question of how a machine finds a service it has never seen and pays for it in the same breath. Orthogonal picked that boring, load-bearing question on purpose - and got a serious set of investors to agree it's worth answering.