Naïve Raises $28.5M to Let AI Agents Build and Run Entire Companies

Palo Alto-based Naïve Inc. announced on August 6, 2026 that it has raised $28.5 million in a Series A round led by Nexus Venture Partners, with Y Combinator, Zetta, and Liquid 2 joining. The company builds infrastructure that lets AI agents handle nearly every step of starting and running a business, from company incorporation and cloud setup to payments and accounting, all through a single API. It already has more than 30,000 developer customers and says revenue grew more than tenfold in the past six months.

What happened

Detail	Fact
Round size	$28.5 million Series A
Lead investor	Nexus Venture Partners
Other investors	Y Combinator, Zetta, Liquid 2, angel investors
Developer customers	30,000+
Revenue growth (last 6 months)	More than 10x
Annual revenue run-rate	Low double-digit millions
Headquarters	Palo Alto, California

Naïve provides an API that third-party AI coding agents, including Claude Code, Codex, and Cursor, call to provision everything a new business needs. Developers give the agent a prompt and a token budget; the agent connects to Naïve’s API and handles the rest.

The infrastructure it can set up includes: US limited liability company incorporation (with state, business description, industry code, and proposed name), email inboxes, phone numbers, databases, cloud compute, storage, and connections to payment processors like Stripe and accounting platforms like QuickBooks.

There are two hard stops where a human must still act. The business owner is required to complete Know Your Customer (KYC) and Know Your Business (KYB) verification, and to make any required payments, themselves. Everything else, according to the company, can be handed off to agents.

What does Naïve actually build for cost control?

Running a swarm of AI agents is expensive. Agents make repeated calls to large language models, pass long context windows back and forth, and burn compute even when idle. Naïve has built three things to address this directly:

1. Model router: directs each task to the most cost-effective model available rather than defaulting to the most capable one.
2. Memory system: stores and reuses business context so agents do not re-process the same information repeatedly.
3. Serverless runtime: runs agents in lightweight JavaScript environments to reduce idle compute costs.

CEO and co-founder Sean Dorje framed this as the core long-term bet: “Our vision at Naïve is to make each token do more, so autonomous companies can become a cost-efficient reality.”

The platform also includes a governance layer where operators can cap budgets, restrict agent capabilities, and require human approval before specific actions are taken.

Why it matters

The pitch from Nexus Venture Partners investor Abishek Sharma is that autonomous software is already here, and fully autonomous companies are the next step. Whether or not that framing holds, the underlying product is real: customers have already built AI-managed TikTok and YouTube content channels, and fully agent-run car rental firms using the platform.

For business owners already thinking about workflow automation, Naïve represents a significant jump in ambition. Most automation tools handle a single workflow in isolation. Naïve is trying to connect company formation, infrastructure, operations, and ongoing cost management into one programmable surface.

The templates on offer go well beyond setup tasks. Naïve includes templates for AI-assisted SEO, full-stack SaaS applications, customer support, recruiting, and accounting. There is even a mobile emulator so agents can navigate smartphone apps on virtual devices.

Our take

The fundraise and revenue numbers are credible. Tenfold growth in six months from 30,000 developers is a meaningful signal that there is genuine demand, not just curiosity. The KYC/KYB carve-out is also honest: regulators require a human in that loop, and any platform that claimed to fully automate it would be lying.

The more interesting play here is not incorporation, it is the cost layer. Any business running agents at scale will quickly discover that token costs compound fast. A model router and memory system that actually work would have retention value far beyond the initial setup. That is probably what Nexus and Y Combinator are betting on.

The risk is complexity. Stitching together incorporation, Stripe, QuickBooks, cloud compute, and phone numbers through a single API is a lot of integration surface to keep stable. One broken third-party integration breaks the whole promise. Teams considering this for anything mission-critical should treat it as early infrastructure, worth watching and testing, but not yet the only layer between you and your business operations.

If you are building automations that need to bridge multiple business systems today, it is worth comparing what Naïve offers against what you can already assemble through AI integration work with existing tools before committing to a new platform dependency.

What to do about it

1. Check whether your use case is primarily setup (one-time) or ongoing operations (recurring), since the ongoing cost-reduction features are where the real value sits.
2. Review Naïve’s API documentation to see which third-party services it currently supports and whether your stack is on that list.
3. Test the model router in a sandboxed project before using it on anything that touches real customer data or money.
4. Map which actions in your agent workflows genuinely need human approval and configure the governance layer before any production deployment.

The setup story is compelling on paper. The cost-efficiency story is what will determine whether Naïve becomes infrastructure or a curiosity.

Building an automation like this? Most client workflows we ship run on Make (referral link, it supports our reporting). If you would rather have it built and monitored for you, that is our workflow automation service.

Source: Bing News · Make.com