Naïve Raises $28.5M

Paris-based startup Naïve closed a $28.5 million Series A on August 6, 2026, led by Balderton Capital, to fund a platform that lets AI agents handle the entire administrative lifecycle of running a company — from filing the LLC to setting up cloud infrastructure — through a single API call. The company has already attracted more than 30,000 developer sign-ups, and what started as a business-formation convenience tool has quietly pivoted into something more technically interesting: infrastructure for making AI agent fleets cheaper to operate.

What's Converging

The last eighteen months of the software industry can be summarized as a slow erosion of the line between "building a product" and "running a company." Developers using tools like Cursor, Claude Code, and OpenAI's Codex can now ship production-grade applications by describing what they want in plain English. The phrase "vibe-coding" — building software through intent rather than syntax — went from a joke to a genuine workflow pattern fast enough that entire categories of developer tooling have had to rethink their positioning. The bottleneck for solo founders and small teams has shifted away from writing code entirely, and onto the unglamorous operational stack underneath it: incorporation paperwork, payment processors, tax accounts, phone numbers, legal agreements.

That operational gap has drawn increasing investor attention. Multiple reports from the first half of 2026 point to a cluster of startups targeting what one analysis calls the "post-vibe-coding problem" — once you have a working product built by AI, who runs the company? The AndroGuider writeup on Naïve's raise frames the company explicitly as "the operational layer for the vibe-coding movement," noting that its founders came from Revolut and Stripe — organizations that spent years automating exactly this kind of financial plumbing at enterprise scale. That background matters because the failure mode most observers expect in this space isn't a lack of AI capability; it's the legal and financial liability that comes from automating high-stakes decisions through systems that can hallucinate.

At the same time, a separate trend is pulling in the same direction from the infrastructure side. The cost of running a fleet of AI agents — paying for inference across hundreds or thousands of concurrent calls — has become a meaningful line item for any serious autonomous product. Companies building on top of foundation models have started treating model routing, cost optimization, and context caching not as engineering nice-to-haves but as product differentiators. Naïve's fundraise lands squarely at the intersection of both pressures.

The Specific Development

Naïve's pitch, at its simplest, is that starting a business has become just another workflow worth scripting. Hand an AI agent a prompt, and the platform provisions the legal entity, opens bank accounts, sets up email inboxes, connects payment processors like Stripe and QuickBooks, allocates cloud storage and compute, and assigns virtual cards and phone numbers — all through one API. Founders still clear KYC and KYB checks themselves, and payment approvals stay human-gated, but the rest of the setup stack is handled programmatically. The company's reported examples of what customers are building on top of it include autonomous rental car agencies, faceless TikTok content operations, and AI automation agencies — businesses designed from the start to run with minimal human involvement.

The $28.5M round, with participation from existing investors and fintech and legal-tech angels alongside lead investor Balderton Capital, reportedly values the company in a range consistent with a 10x ARR run in the six months preceding the raise — achieved with a team of ten people. That ratio is the number that deserves attention, not the headline funding figure.

Our read is that the more consequential part of this story isn't the business-formation toolkit — it's what Naïve calls its agent infrastructure layer. The company has built a model router that doesn't just select the cheapest available model for a given task; it caches already-reasoned outputs and replays them for downstream agent calls that would otherwise re-derive the same conclusions from scratch. Combined with a contextual memory system and a serverless JavaScript runtime that charges only when agents are active, the infrastructure stack functions as a cost-containment layer for anyone running long-lived or high-frequency autonomous agents. Multiple reports indicate that the fastest-growing demand among Naïve's 30,000 developer adopters isn't the setup tooling — it's inference cost reduction. That inversion tells you what the real product is.

The governance layer deserves its own mention. Budget caps, capability restrictions, and human-approval gates are exposed as first-class API primitives, not afterthoughts. The fact that 30,000 developers adopted this voluntarily — rather than being forced into it by compliance requirements — suggests the market understands that fully autonomous agent fleets without control knobs are operationally unsafe, and will pay for a clean abstraction over those controls.

What's Likely Next

The most immediate question is whether the legal and compliance surface area can hold up at scale. The AndroGuider analysis flags what it calls the "garbage in, garbage out" problem directly: AI systems handling high-stakes financial and legal decisions across multiple jurisdictions are only as reliable as the workflows underneath them. Naïve's human-in-the-loop checkpoints on payments and KYC are a real constraint, but they are also a narrow gate. What happens when a developer runs an autonomous business in a jurisdiction with tax rules the platform's deterministic workflow engine wasn't trained on? Cross-jurisdictional compliance is the wall this category of product hits first, and how Naïve's team — with their Revolut and Stripe backgrounds — handles it will determine whether the platform can move upmarket beyond solo founders and small teams.

The second watch item is whether the reasoning-replay router pattern gets productized and licensed independently of the business-formation layer. If inference cost reduction is genuinely the fastest-growing demand signal, there is a credible argument that the agent infrastructure stack is the business, and the LLC-in-a-prompt feature is a compelling acquisition channel that got them there first. The $28.5M gives the team enough runway to find out which product the market actually wants to pay for. Within the next 90 days, watch for whether Naïve begins positioning its model router and serverless runtime as standalone offerings — or whether it doubles down on the full-stack autonomous business narrative and tries to land its first enterprise customer willing to run a real operation on it.

Sources