Naïve secures $28.5M to build infrastructure for AI-run businesses
Naïve has raised $28.5 million in a Series A led by Nexus Venture Partners, following the rapid adoption of its infrastructure for AI agents that set up and operate business services. The startup says it signed more than 30,000 developer customers within months of launch and increased annual run-rate revenue tenfold to the low double-digit millions during the past six months.
The company packages payments, email accounts, phone numbers, cloud infrastructure, storage and company incorporation behind a single API. Developers can give Naïve’s prompt to tools including Cursor, Claude Code and Codex, enabling those tools to connect to its APIs and provision the underlying services.
Business setup through agent infrastructure
Naïve says an agent can orchestrate the formation of a U.S. LLC by supplying information such as the state, industry code, business description and proposed names. Users must still participate in KYC and KYB processes and make required payments.
After those steps, AI can handle tasks including creating email inboxes, virtual cards, phone numbers, databases and computing resources, as well as connections to Stripe and QuickBooks. The company also offers templates for AI SEO, full-stack SaaS applications, recruiting, accounting, customer support and a mobile emulator that lets agents operate smartphone apps on emulated devices.
A governance layer is designed to let customers set budgets, restrict what agents can do and require human approval before sensitive actions. This focus on controls accompanies customer use cases that CEO and co-founder Sean Dorje said include AI automation agencies, anonymous TikTok and YouTube content channels, and a rental-car agency.
Lowering the cost of running agents
Naïve is directing part of the new funding towards infrastructure intended to make agent loops more efficient. Its planned projects include a model router that selects an efficient model for each task while preserving and replaying previously reasoned data, a memory system for storing and surfacing business context, and an orchestrator that divides work among agents.
The startup is also developing a serverless runtime that runs agents in lightweight JavaScript environments rather than giving each agent a full virtual machine. Naïve says this approach means customers pay primarily when an agent is active and can reduce the cost of deploying many agents.
The infrastructure emphasis fits a wider market in which AI compute capacity is becoming a strategic concern, as accelerating neocloud market for AI reflects the accelerating neocloud market for AI. Dorje said inference and serverless agents are among Naïve’s fastest-growing areas of demand, and that the business has drawn enterprise interest.
Funding plans and operational implications
Naïve has 10 full-time employees. The Series A also included Y Combinator, Zetta, Liquid 2, Gokul Rajaram, Apollo.io co-founder Tim Zheng and former HubSpot COO JD Sherman. The company plans to hire researchers and advance four infrastructure areas: virtualized agent sandboxes, model routing and inference optimization, a memory layer, and governance and orchestration.
For businesses assessing autonomous workflows, Naïve’s proposition is not only faster provisioning of operational tools but also closer attention to the continuing cost, permissions and supervision of agents once they are in production.
What Naïve’s Series A means for AI agent infrastructure
The announcement illustrates a broader shift from standalone AI tools towards infrastructure that can provision services and coordinate business tasks. The practical question is not only what an agent can launch, but how its access, costs and actions remain controlled in production.
From rapid setup to controlled operations
Naïve brings payments, email accounts, phone numbers, cloud resources, storage and company-incorporation services behind one API. However, the report also notes that users still participate in KYC and KYB checks and complete required payments. This distinction matters: automation can coordinate operational steps without removing human responsibility for sensitive decisions.
- Human approval remains relevant for sensitive actions.
- Budgets and permissions should be defined before an agent runs.
- Memory and orchestration affect how agents use business context.
- Operating costs should be assessed alongside setup speed.
What businesses should evaluate
Agent infrastructure should be assessed as an operating model, not only as a faster way to configure software. Before production use, a business should map which systems an agent may access, which actions need approval, what context may be retained and how activity will be reviewed.
- Define clear boundaries for each connected system.
- Require approval where an action creates financial or operational risk.
- Review how stored context is retained and surfaced.
- Monitor usage as agents move between models and external services.
Frequently asked questions
How much did Naïve raise in the Series A?
The funding figure reported on this page is $28.5 million, with Nexus Venture Partners leading the round.
What does the unusual search fragment “32miyertechcrunch” mean here?
That fused fragment appears in a search-query variant, but it is not treated as a funding figure or source attribution. The supported details in this report are a $28.5 million Series A led by Nexus Venture Partners.
Does this mean AI agents can establish and operate a business without people?
No. The described infrastructure can orchestrate many setup and operating tasks, but users still participate in KYC and KYB processes, make required payments and can require human approval for sensitive actions.