Today, we’re announcing Naïve has raised a $28.5 million Series A to build the infrastructure AI agents need to run entire businesses.
This round was led by Nexus Venture Partners, with participation from Y Combinator, Zetta, Liquid 2, and more.
Co-Founders Sean Dorje & Dennis Zax have been building together since age 14. They sold their first company as teenagers, dropped out of Berkeley, joined Y Combinator, and spent the last several years thinking about what the world looks like when agents can fully run businesses.
The Problem: The App Ships in an Afternoon. The Business Takes Months.
Coding agents can now ship a working app in an afternoon. But turning that app into a real operating business still takes months.
Developers manually stitch together payments, incorporation, email, phone numbers, compute, memory, databases, authentication, and dozens of other services. All of this infrastructure built for human operators, not autonomous agents.
The Solution: One Config File, One API, the Entire Operating Stack
Naïve is rebuilding and unifying the entire business stack for agents.
Today, thousands of developers write a single config file, and Naïve provisions everything their agents need behind one unified API:
Real-world Identity: KYC/KYB, LLC formation, Email, Phone, Virtual Cards Cloud: Database, Hosting, Compute, Storage, Auth Serverless Agents: Hermes, Claude Code, Codex, Vetta Orchestration: Task Management, Interagent Communication, Scheduling Inference / Model Routing Memory: One context layer that grows with the entire company. 10k+ Tool Gateway: Stripe, Github, Quickbooks, Higgsfield, Browser, etc. Governance: Enforce capability policies, audit logs, and HITL approval
The Future: Multiplying Intelligence Per Token
With this new capital, Naïve is building a lab around a simple observation: autonomous companies increasingly rely on agent loops to finish work. However, these agent loops are extremely inefficient due to frontier model lock-in, idle compute costs, and poor context management. Our vision is to maximize the work per token spent on these agent loops through 4 areas:
- Serverless Agents.
minimize compute. Agents and their sandboxes run as V8 isolates. No VMs, no rented boxes sitting idle. Since agents spend 95% of their time waiting, you're only billed for the milliseconds they're actually executing. 2.3ms cold start, 1.2MB per agent.
- Inference.
minimize cost per token. Every call routed to the cheapest model that can answer it, small language models for mechanical turns, batched inference for long-running agents.
- Brain.
minimize input tokens. One memory layer for the company. Distills conversations into structured facts and hands each agent only the context that call needs. Beats state of the art on recall at ~11x fewer tokens per query.
- Agent Workforce Orchestration.
minimize wasted turns. Delegates work to the right subagent at the right tier instead of running everything at frontier cost. Hosted serverlessly at 1/100th the cost.
Our early benchmarks can start an agent in 2.3 milliseconds using 1.2MB, reduce inference costs several-fold, and achieve stronger memory recall while using 11x fewer tokens per query.
In this next decade, agent spend is set to scale into the trillions as companies completely run on agents that finish work autonomously. Our vision with our research is to make each token do more, so autonomous companies can become a cost-efficient reality.
Thank you to all our investors including:
Abhishek Sharma (Partner @ Nexus)
Andrew Miklas, Diana Hu (Partners @ YC)
Apoorva Pandhi (GP @ Zetta VP)
Raj Singh (GP @ Z21 Ventures)
Jimmy Cho (Investor, DEEPCORE (Softbank))
Michael Ma (Partner, Liquid 2 Ventures)
Vijay Krishnan (Co-Founder of Turing)
Gokul Rajaram
JD Sherman (Former Hubspot President)
Tim Zheng (Founder of Apollo)
Robert Chatwani (President @ Docusign)
Zachary Sims (Founder of Codeacademy)
Gert Lanckriet (Head of ML @ Amazon)
Eric Ciarla (Co-Founder of Firecrawl)
We’re now hiring benchmark breakers who want to push the frontier of agent business infrastructure.
Learn more and start building at usenaive.ai.
