Nous Research just raised $90 million to turn a free, community-built agent into a business product, and the agent in question has already been cloned more than 24 million times, according to the company. That single fact says a lot about where open source AI agents for business are heading in late 2026. Teams that once assumed agents had to come from a handful of closed vendors now have a credible alternative they can inspect, host, and shape themselves.

The company behind it is Nous Research, and the product is Hermes Agent, which is about to get an enterprise edition called Hermes for Businesses. Below, we break down what was announced, what the open source approach changes for builders, how to evaluate an open agent for your own stack, and the risks worth watching. If you build with or buy autonomous AI, this is a story to follow closely.

What Nous Research Announced and Why It Matters

On October 7, 2026, Nous Research announced a $90 million Series B to expand Hermes Agent for business customers. The Wall Street Journal reported a valuation of $1.5 billion, a figure Nous itself did not disclose. The Next Web reported that investors include Nvidia, Microsoft’s M12, Samsung, Union Square Ventures, Y Combinator, and Menlo Ventures.

Hermes Agent was released in February 2026 under the permissive MIT license. It remembers projects, creates reusable skills as it works, and lets users pick which model powers it. It runs on macOS, Windows, and Linux, and people can reach it through Telegram, Slack, and WhatsApp. Nous says it has been cloned more than 24 million times and estimates it drives roughly 2.5 percent of global token usage, though those are company figures, not independent audits.

The new enterprise product targets organizations that want to choose models based on price and goal, keep their AI usage private, and own their internal knowledge. That last point is the heart of the pitch. When your agent’s memory and skills live in your own environment, you are not renting your institutional knowledge back from a vendor.

Open Source AI Agents vs Closed Platforms

The agent market is splitting into two camps. On one side are closed, vendor-run agents such as the coworker tools we covered in our guide to AI coworker agents and Google’s Gemini Agent. On the other are open agents you can run, fork, and extend. Each has real strengths.

Closed platforms usually win on polish, support, and tight integration with the suite you already pay for. Open agents win on flexibility. Because Hermes lets you swap the underlying model, a team can route cheap tasks to a small model and reserve a frontier model for hard reasoning. That model-agnostic design also protects you from a single provider’s pricing changes or outages.

Open source also changes the security conversation. You can read the code, pin a version, and decide exactly which tools and data the agent touches. That does not make it automatically safe. It makes it auditable, which is a different and more useful property. If you want a refresher on why guardrails matter, our breakdown of AI agent security and the lessons from rogue agents is a good companion read.

Finally, there is cost. Open agents let you negotiate with the model market rather than with one vendor, and for high-volume workflows that flexibility can matter more than any single feature.

How Do AI Agents Automate Business Tasks in Practice?

If you are wondering how AI agents automate business tasks, the Hermes design offers a clear picture. The agent receives a goal through a chat channel your team already uses, such as Slack. It plans the steps, calls tools, and records what worked as a reusable skill. The next time a similar request arrives, it reuses that skill instead of starting from zero.

That learning loop is what separates a durable agent from a one-off chatbot. A support team might teach it a refund triage routine once. A finance team might teach it a month-end reconciliation checklist. Over weeks, the agent accumulates a library of procedures specific to your company.

Here is a practical way to evaluate any open source agent before committing:

• Start with one contained workflow. Pick a repetitive task with clear success criteria and low risk if it goes wrong.
• Check the license and the roadmap. MIT is permissive, but confirm what the enterprise edition adds and what stays free.
• Map data access explicitly. Decide which systems the agent can read and which it can write to.
• Keep a human in the loop. Require approval for irreversible actions such as sending money or deleting records.
• Measure cost per completed task. Track tokens, retries, and human review time, not just model price.

Long-running agents add another layer of planning, which we explored in our explainer on long-running AI agents and Microsoft Autopilot.

The Future of Enterprise Agents: Open, Hybrid, and Accountable

Expect most companies to land on a hybrid model. They will use closed agents where convenience is king and open agents where control, privacy, or cost matters most. The funding round suggests investors believe the open side can support a real business, not just a hobbyist community.

There is a fair counterpoint. Open source does not remove operational burden. Someone still has to host the agent, patch it, monitor it, and answer for its mistakes. Download counts also tell us about curiosity, not production success, and a clone is not a deployment. Hermes for Businesses will be judged by how well it handles governance, audit trails, and support, the unglamorous features enterprises actually buy.

Standards may be the next battleground. As agents from different vendors need to talk to each other, shared protocols could matter as much as any single model. Teams that stay model-agnostic today will be best positioned to adopt them later.

Key Takeaways and Next Steps

First, open source AI agents for business are now backed by serious capital, which makes them a safer bet to evaluate. Second, model flexibility and data ownership are the real advantages, so design your pilots around them. Third, openness does not replace governance; keep humans in the loop and define clear boundaries.

Want more practical coverage of agent tools, trends, and strategies? Explore BigAIAgent for the latest articles and resources.

So here is the question: would you trust an agent you can fully inspect and host yourself more than one you rent from a vendor, and why?