When Sean Dorje and Dennis Zax set out to build infrastructure for autonomous AI agents, they anticipated developers would need help with compute and payments. What they didn't fully anticipate was how quickly those developers would also want their agents to open bank accounts, form LLCs, and post to TikTok.
That's the territory Naïve has staked out—and investors are paying attention. The San Francisco startup announced a $28.5 million Series A on August 6, 2026, led by Nexus Venture Partners, with participation from Y Combinator, Zetta, and Liquid 2, according to TechCrunch. Angel checks came from a roster that reads like a Silicon Valley Rolodex: Gokul Rajaram, Apollo.io cofounder Tim Zheng, former HubSpot COO JD Sherman. Total capital raised now sits around $32 million.
For a company that went through YC's Spring 2025 batch—barely a year ago—that's a compressed timeline. Then again, the traction numbers tell their own story.
Scaling Faster Than Most Expected
Naïve claims it hit 30,000-plus developer customers within months of launch, according to the August announcement. More striking, perhaps: the company says its annual recurring revenue scaled tenfold over the past six months, reaching what it describes as the low double-digit millions. The startup is doing this with around 10 full-time employees.
Those metrics, if they hold, would put Naïve in rarefied company among infrastructure startups. Whether that growth sustains is the bet Nexus and its co-investors are making.
What Exactly Is Naïve Building?
The pitch centers on what the company calls "unified agent infrastructure"—a single API handling everything an autonomous AI agent might need to operate in the world. Not just thinking and acting, but existing as an entity with identity and transactional capacity.
Developers get access to a catalog of more than 40 "primitives," API-accessible building blocks spanning company formation, banking and cards, email and domains, phone and SMS, image and video generation, social media distribution, and data intelligence. The platform routes across 300-plus AI models and integrates with the usual suspects: Stripe, Supabase, Vercel, QuickBooks, Mercury, Rippling.
In practice? A developer can invoke Naïve's /formation primitive to spin up a U.S. LLC, /kyc for identity verification, /cards to issue virtual cards, or /social to post across 10 platforms—all through a unified config file or API. It's infrastructure as bureaucracy-solver, essentially. Agents that can actually do things in the real economy, not just chat.
The platform includes governance features—spend caps, approval workflows, audit logs—meant to keep agents from going rogue with a corporate credit card. Pricing runs on a credit system: each credit costs five cents.
Naïve operates as a product of Relixir, Inc., according to company filings. Dorje serves as CEO; Zax as CTO.
The Build-Out Ahead
So what does a $28.5 million round buy? Naïve outlined four infrastructure projects: virtualized sandboxes where agents can run safely, model routing and inference optimization to cut costs, a memory layer for agents, and enhanced governance and orchestration capabilities. The company is also shipping a serverless runtime for JavaScript-based agents, designed to minimize costs when agents sit idle.
It's a roadmap that assumes agents will soon need the same kinds of operational scaffolding that human organizations rely on. Whether that assumption proves prescient or premature will depend, in large part, on how quickly autonomous agents move from prototype to production.
Crowded Territory
Naïve's announcement lands in the middle of what might generously be called an agent infrastructure arms race. Prime Intellect raised $130 million at a $1 billion valuation in July 2026. CopilotKit secured $27 million in May. Gumloop landed $50 million from Benchmark in March. Meanwhile, OpenAI, Anthropic, and Google have all pushed updates this year aimed at making their models more agent-capable.
The underlying thesis is the same across most of these bets: if AI agents are going to operate autonomously at scale, they'll need infrastructure that doesn't exist yet. Banking rails. Legal scaffolding. Identity systems. Someone has to build it.
Naïve is betting that "someone" might as well be them.
Whether the market for autonomous agents materializes at the scale investors are banking on remains an open question. But for now, at least, the capital is flowing—and Dorje and Zax have a 10-person team to prove the thesis.
