Forming a company, setting up an email account, opening a bank account, getting a corporate card, spinning up cloud servers — the grunt work every founder dreads. A startup just made all of it something you hand off to an AI coding agent with a single prompt.

3-second summary
One prompt AI forms a US LLC Bank account + card issued Cloud + payments auto-configured An unmanned company goes live

Wait, you can hand an entire company to AI?

San Francisco-based AI infrastructure startup Naïve just raised a $28.5M Series A. Nexus Venture Partners led the round, joined by Y Combinator, Zetta, Liquid 2, and angels including Apollo.io co-founder Tim Zheng and former HubSpot COO JD Sherman. Total funding to date is around $32M.

The product itself is simple. A developer drops a prompt into Cursor, Claude Code, or Codex, and the coding agent calls Naïve's API to provision payment methods, an email account, a phone number, cloud infrastructure, storage, and a US LLC — all at once. The only things a human still has to do are identity checks (KYC/KYB) and required payments.

The founders are 20-year-old Berkeley dropouts Sean Dorje (CEO) and Dennis Zax (CTO), who've been building together since they were 14. They sold their first startup, ezML, as teenagers before joining YC's Spring 2025 batch to start Naïve.

Within months of launch, it had signed up over 30,000 developer customers, and annualized revenue has grown 10x in six months, into the low double-digit millions. The fastest-growing use case is AI automation agencies, followed by "faceless" AI-generated YouTube and TikTok channels — and one customer is running an entire car rental agency with no human operator at all.

So how's this different from Stripe Atlas?

Stripe Atlas, Firstbase, Clerky, Doola — names every founder knows. All of them are built around a human filling out a web form. Banking is a separate application, and the whole process can take days to weeks.

Agent-native formation works differently. There's already a public walkthrough showing a clean Delaware LLC — funded bank account included — going from prompt to done inside a single conversation, in about 15 minutes. What sets Naïve apart is that it doesn't stop at incorporation — it bundles everything the company needs to keep running afterward: email, payments, cloud, tool sign-ups, all under one API the agent keeps operating against.

Traditional formation servicesAgent-native infrastructure
How you applyHuman fills out a web formPrompt to a coding agent
Bank accountSeparate application, waiting on approvalOpens automatically with formation
After formationHuman sets up email, payments, infraAgent keeps running operations
Spend controlsN/A — a human is payingBuilt-in budget caps + approval gates

Okay, but doesn't this go wrong somehow?

Fair worry. Back in May, an AI agent built by a separate agent-infrastructure project called ClawBank made headlines after it autonomously obtained a US IRS Employer Identification Number (EIN) and an FDIC-insured bank account on its own. The agent ran an X account under the pseudonym "Manfred Macx" — named after a character from a sci-fi novel — and got itself set up to trade over 30 cryptocurrencies.

It was framed as an experiment, but the lesson is clear: the moment an AI agent gets a bank account and spending authority, money can move somewhere no human is watching — unless there are controls in place. That's exactly why Naïve baked a governance layer into the product from day one: budget limits, restricted agent capabilities, and human approval required before sensitive actions.

Worth remembering

The legal owner is still a human. The AI is just an agent — the LLC itself, and the liability that comes with it, belongs to the human founder who passed KYC/KYB.

Alright, how would I actually start?

  1. Decide what you're actually handing off first
    Pick one slice — incorporation, email, payments — instead of everything at once. Start with a single workflow, like the car rental example, before you expand scope.
  2. Write your governance rules before you write any prompts
    Set budget caps, restrict what the agent can do, and decide which actions need a human sign-off. Handing over a wallet with no guardrails, like the Manfred case, is hard to walk back.
  3. Put legal ownership in writing
    KYC/KYB and the LLC itself belong to a human. "The AI did it" won't hold up as a legal defense later.
  4. Pilot small before you scale
    Test in a low-stakes area first — a faceless content channel, say — and expand once it's proven out.
  5. Lock in a review cadence
    Put a recurring check on your calendar, weekly or monthly, to actually look at spend and activity logs. A governance layer nobody checks isn't doing anything.

Want to dig deeper?

Naïve raises $28.5M The original TechCrunch report on the funding, with product details and founder interviews techcrunch.com

Naïve on Y Combinator Company overview, founder info, and YC batch record ycombinator.com

Investor commentary in full More on what Nexus Venture Partners and the CEO had to say siliconangle.com

The first AI agent to form its own company How ClawBank's Manfred agent got an EIN and a bank account coindesk.com

Stripe Atlas alternatives, 2026 A guide comparing traditional LLC formation services against agent-driven ones meow.com

Naïve's founder story A rundown of the 20-year-old Berkeley dropouts' prior startup history thesaasnews.com