Naive raises $28.5 million to build infrastructure for AI-run businesses

Naive has raised $28.5 million in a Series A round led by Nexus Venture Partners to expand infrastructure designed for businesses operated with AI agents. The financing was reported on August 6 after the company signed up more than 30,000 developer customers. Naive packages services needed to establish and run a business behind an API, allowing agents to coordinate resources that would otherwise require separate accounts and setup processes. The company is presenting itself not as the source of the agents, but as the operational layer they can use to assemble payments, communications, computing, storage and company formation.

The service connects to coding agents through a prompt and API. According to the company, an agent can supply details for forming a United States limited liability company, including its proposed names, business description, state and industry code. Human participation remains necessary for identity and business verification and for required payments. That distinction limits the claim of autonomy: Naive can orchestrate much of the setup, but people retain responsibility at regulated or financially consequential checkpoints. Other supported tasks include creating inboxes, phone numbers, virtual cards, databases and computing resources, along with connections to services such as Stripe and QuickBooks.

Naive also offers a governance layer through which users can set budgets, restrict what agents are allowed to do and require approval before sensitive actions. Templates cover use cases including software services, recruiting, accounting and customer support, while a mobile emulator lets agents operate smartphone apps through emulated devices. These pieces reflect the breadth of the company's pitch. It is trying to make the fragmented infrastructure of a business available through machine-readable controls, then place limits around how automated systems use it. The result is closer to an operating toolkit than a single autonomous-company application.

The startup says demand has translated into rapid financial growth. Chief executive and co-founder Sean Dorje told TechCrunch that annual run-rate revenue increased tenfold during the previous six months to the low double-digit millions. He said customers were using Naive for AI automation agencies, online content operations and even a rental-car agency. Those are company-reported examples and metrics, but they help explain the size and timing of the Series A. The funding round also included Y Combinator, Zetta, Liquid 2 and several angel investors, bringing Naive's total capital raised to roughly $32 million.

The newly raised money is intended to support four infrastructure initiatives rather than mark four completed product launches. Naive plans virtualized sandboxes for agents, model routing and inference optimization, a memory layer, and governance and orchestration. The model router is meant to direct a request toward an efficient model while preserving work that has already been reasoned through. The memory system is intended to store business context and surface it when agents need it, while orchestration would divide work among multiple agents. These projects address the recurring operation of agents after a company has been created, not merely the initial setup process.

Cost is central to that expansion. Agents can repeatedly call models, carry large contexts and consume infrastructure even when their useful work is intermittent. Naive is building a serverless runtime that would place agents in lightweight JavaScript environments rather than dedicate a complete virtual machine to each one. The company says this design would let customers pay mainly while an agent is active and make large deployments less expensive. Dorje said inference and serverless agents are among the fastest-growing areas of demand, including interest from enterprises, although he did not name prospective enterprise customers.

Naive currently has 10 full-time employees and plans to use part of the round to hire researchers as it develops the four initiatives. The financing validates investor interest in an emerging layer beneath agent-run services, but execution remains the next test. Provisioning a phone number or database is a discrete operation; maintaining memory, controlling permissions and routing model calls across a continuously operating business is a more complex systems problem. Naive has completed the fundraising and reports strong early traction. Its broader infrastructure roadmap, however, should be judged as work funded by the round rather than technology that the August 6 report established as already fully delivered.