Nous Research Builds an Open-Source Enterprise AI Agent

Nous Research builds open-source AI systems for people and organizations that want powerful agents without surrendering every decision, model choice, and piece of institutional memory to one closed platform. Its operating wager is that user control can become an advantage rather than a tax.

Formed in 2023 by co-founders Jeffrey Quesnelle, Karan Malhotra, Ryan Teknium, and Shivani Mitra, the Austin company is led by CEO Dillon Rolnick. Its best-known product is Hermes Agent, MIT-licensed software that can run on user-controlled infrastructure, work across models, use tools, remember projects, and turn experience into reusable skills.

Nous Research matters now because open-source popularity is no longer the finish line. The company is trying to turn a widely distributed developer product into enterprise infrastructure, where security reviews, permissions, support, and accountability arrive before the applause.

About Nous Research

Nous Research was created around a blunt premise: advanced AI should not become a gated utility controlled by a few providers. The company develops open models, post-training methods, agent software, and infrastructure intended to make AI development and operation more accessible.

That mission has produced more than a model catalog. Nous developed DisTrO and the Psyche network to explore training across distributed hardware with far less communication overhead than conventional approaches. It also built Hermes Agent as a practical layer for running tools, maintaining memory, communicating across interfaces, and choosing among model providers.

The throughline is control. Nous wants builders to choose where an agent runs, which model it uses, what it remembers, and how it evolves. That sounds philosophical until an organization has to decide whether its workflows and knowledge can ever leave a vendor's platform.

Hermes Agent turns openness into a product decision

Hermes Agent is open-source under the MIT license. The software is designed to run on anything from modest servers to larger infrastructure, connect to different model providers, use a broad toolset, retain memory across sessions, and create reusable skills from completed work.

Nous describes Hermes as a self-improving agent with a built-in learning loop. The practical value is not that software magically escapes supervision. It is that an operator can inspect the system, choose the deployment environment, change models, extend capabilities, and keep the operating context closer to the organization using it.

Nous says Hermes has been cloned more than 24M times and accounts for roughly 2.5% of global AI token use. Those figures are company-reported rather than independently audited, but they indicate a distribution base that most enterprise software companies would happily trade a conference booth for.

The enterprise chapter changes the burden of proof

The Hermes for Businesses strategy is supposed to bring that control into organizations. Nous says companies will be able to customize agents, select models for cost and performance, preserve data control, and build an institutional knowledge layer without locking every workflow to one provider.

That promise creates a harder assignment than winning developers. Enterprises need identity and access controls, audit trails, deployment support, incident response, model governance, documentation, and a clear answer when automated work reaches production and something goes wrong.

Nous now has to make managed service compatible with user control. If the product becomes convenient only by pulling customers back into a closed architecture, the company weakens its own thesis. If it keeps everything portable but leaves buyers to assemble the operational layer themselves, it limits the enterprise market. The durable position sits in the uncomfortable middle.

Leadership is carrying an ideological company into operations

Nous Research's four co-founders, Jeffrey Quesnelle, Karan Malhotra, Ryan Teknium, and Shivani Mitra, established the company during the rapid concentration of frontier AI inside a small group of technology platforms. Current CEO Dillon Rolnick is now responsible for converting that founding argument into an operating company.

The leadership task is unusually specific. Nous does not merely have to ship capable software. It has to prove that open systems can be reliable enough for businesses, flexible enough for researchers, and commercially strong enough to fund continued development.

That is why the company's $90M Series B matters beyond the reported $1.5B valuation. The capital gives Nous time to build security, product, infrastructure, and customer-facing functions around a project that already has meaningful developer distribution.

Hiring shows where the open-source project becomes a supplier

The current Nous Research careers page listed 13 open roles at the time of publication. The company was hiring in security, infrastructure, evaluation, product engineering, mobile development, forward-deployed engineering, analytics, support, product management, finance, and office operations.

That mix says more than a generic hiring banner. A security engineer owns the controls enterprise buyers will examine. A forward-deployed engineer helps adapt Hermes inside customer environments. Support and product analytics turn community use into observable service quality. A product manager for Hermes Business has to package openness without sanding away the control that gives the product its reason to exist.

Nous describes its culture as high-agency and mission-focused, with an unusually direct expectation of intense work. The language will attract some builders and repel others. At least it does not pretend that converting research, open-source software, and enterprise obligations into one company is a casual assignment.

What Nous Research must prove next

Nous Research has distribution, technical ambition, a sharp market position, and enough capital to test the commercial version of its argument. TechCrunch, citing the Wall Street Journal, reported roughly $36M in annualized revenue by mid-September 2026. That is a meaningful signal, though the company's higher year-end target remains a projection rather than achieved performance.

The next evidence will be less theatrical than a valuation. Watch for named enterprise deployments, security and governance capabilities, customer retention, support performance, and clear proof that organizations can move models or infrastructure without losing the knowledge their agents accumulated.

Closed AI platforms sell convenience through concentration. Nous Research is betting that control, portability, and openness can become their own enterprise advantage. Hermes has already traveled through the developer world. The company now has to prove it can walk through procurement without leaving its principles in the lobby.