The $1.5 Billion Bet That AI Agents Should Be Free

Last Tuesday, a company most business owners have never heard of merged 338 pull requests into a single software release. That company, Nous Research, is now worth $1.5 billion.

Their product is called Hermes Agent. It's an open-source AI agent that businesses can use to build autonomous assistants with persistent memory, tool execution, and the ability to learn new skills over time. Think of it as the agent infrastructure that Salesforce and Microsoft charge six figures for, except the code is free.

Nous just closed a $75 million Series B led by Robot Ventures, with Union Square Ventures also in the round. For a product you can download on GitHub right now, that's a lot of investor conviction.

Why this matters if you run a business

The AI agent market has been split into two camps. Enterprise platforms like Salesforce's Agentforce and Microsoft's Copilot charge premium prices for managed, walled-garden deployments. Open-source projects offer flexibility but have lacked the polish, support, and funding to compete seriously.

Hermes Agent just crossed that line.

The September 14 release consolidated months of development into a stable, production-ready version. The product now ships with three deployment options: cloud-hosted tiers starting at $20 per month, on-premises installation, and Docker containers for teams that want full control.

That $20 entry point is the real story. A small marketing agency can now run an AI agent that remembers past conversations, executes multi-step workflows, and improves with use. Six months ago, that capability required a developer team or an enterprise contract.

Persistent memory changes the math

Most AI tools start fresh every conversation. They don't remember what you asked yesterday or what worked last month. Hermes Agent stores context across sessions. It pulls relevant knowledge from past interactions and applies learned skills to new tasks.

This is the gap between a chatbot and a teammate. A chatbot answers questions. An agent with memory handles projects.

For a COO managing operations across departments, this turns AI from a tool you check occasionally into something that actually reduces your team's recurring workload.

What the open-source pressure looks like

When a free product gets $75 million in funding and a $1.5 billion valuation, the paid platforms notice. Salesforce launched Agentforce with named AI agents and enterprise pricing. Microsoft bundles Copilot into existing subscriptions. Both have massive distribution advantages.

But neither can match $20 per month. And neither hands over the source code.

For AI agencies and consultants building client solutions, open-source agent tools mean lower costs, more customization, and zero vendor lock-in. You can deploy Hermes on a client's own servers. You can modify the code. You can swap models without rebuilding from scratch.

What to watch

I've been seeing this pattern in practice: businesses don't want another SaaS subscription. They want AI that runs on their terms, on their servers, with their data staying put.

Nous Research isn't alone in the open-source agent space. But they're the first to combine a mature product, a real revenue model, and serious venture backing. That mix tends to pull in enterprise buyers who were waiting on the sidelines.

If you've been holding off on open-source agents until the tools caught up, the wait is over.

— Mark Garza, Laimen AI