- Nous Research funding talks reportedly target at least $75 million and a $1.5 billion valuation for the Hermes agent developer.
- The Nous Research funding round is said to be led by Robot Ventures, with meaningful participation from USV and other investors.
- Hermes combines local agent automation with built-in skills for search, coding, image understanding, Telegram, and Discord workflows.
- A paid hosted service may give Nous a clearer business model than many open-source AI projects have managed to build.
Table of Contents
Nous Research funding puts a price on the agent race
AI agents have spent the past year sounding like a promise waiting for a product. Now, investors are treating them like a funding thesis. Nous Research funding talks reportedly put the open-source Hermes developer on track to raise at least $75 million at a $1.5 billion valuation, a striking number for a company founded only in 2023.
People familiar with the proposed deal told TechCrunch that Robot Ventures is leading the round, with Union Square Ventures, or USV, taking a significant stake alongside other investors. Nous Research, Robot Ventures, and USV did not comment publicly. A term sheet is not money in the bank, and valuations can move right up until a deal closes. Still, if the reported terms hold, the proposed Nous Research funding signals that investors are making a very specific bet. They think agent software can become a durable business, not another impressive GitHub project with no obvious way to pay the bills.
Nous Research had already raised roughly $70 million before these talks, according to Crunchbase data cited in the report. Its backers include Paradigm, North Island Ventures, OSS Capital, Robot Ventures, and Balaji Srinivasan. Add another $75 million or more, and the Nous Research funding total would give the startup a serious war chest for a field where model inference, cloud compute, security work, and developer support all get expensive in a hurry.
The headline valuation is eye-catching, but the timing may be the more telling detail. The market has moved beyond asking whether a chatbot can draft an email. The question now is whether software can reliably complete a chain of tasks while you are asleep, at work, or simply not interested in babysitting it.
Hermes is built for people who want an agent to do things
Hermes arrived shortly after OpenClaw caught fire among developers and AI hobbyists. OpenClaw’s appeal is straightforward: run an agent on a local computer, give it access to tools, and let it take actions on your behalf. The appeal is also slightly chaotic. Anyone who has tried to configure a local agent knows the experience can feel less like installing an app and more like assembling furniture after losing the instructions.
Hermes tries to reduce some of that friction with built-in skills, including web search, coding, and image understanding. It is also designed to learn from usage and develop additional skills without users manually wiring every capability together. That is an ambitious pitch, and frankly, it is where agent companies earn either loyalty or distrust. An agent that improves its toolkit sounds useful; an agent that makes opaque decisions with access to your files and accounts sounds like a problem waiting for a bad Tuesday.
The company has also released language models aimed at coding and mathematics, giving Hermes more of a homegrown technical stack than a thin wrapper around someone else’s API. Users can communicate with agents through Telegram and Discord, run them from a desktop machine or virtual private server, and have them perform automated work remotely around the clock.
That always-on angle is central to the Nous Research funding story. A normal chatbot waits for a prompt. An agent is supposed to notice, plan, call tools, report back, and keep moving. The reported Nous Research funding would help determine whether that promise can be turned into a dependable product. In practice, the difference is like hiring a junior assistant versus opening a search box. The first could save real time. It could also book the wrong meeting, run an expensive cloud job, or confidently pursue an utterly wrong plan.
Hermes has roughly 214,000 GitHub stars and nearly 40,000 forks. GitHub stars are not revenue, and they are definitely not proof that every user is active. But those figures do signal intense developer attention. Forks are particularly meaningful because they suggest people are taking the project apart, adapting it, and trying to make it their own.
Open source brings distribution, then the hard part starts
The reported Nous Research funding valuation points to a pattern that has become familiar in AI: open source can create distribution at a pace traditional software marketing cannot match. Developers discover a tool, install it, share workflows, fix bugs, and build a community before a sales team has even booked its first demo. Meta used a version of this playbook with Llama, while Hugging Face built an ecosystem around making models easier to find and use.
But agent software adds a complication. These systems do not merely generate text or images; they can access browsers, code repositories, messaging apps, servers, and other tools. That raises the bar for reliability and safety. A hallucinated sentence is annoying. A hallucinated command that deletes a production database is a career event.
For Nous, the hosted Hermes offering may be the commercial answer. The company sells cloud access in tiers reportedly ranging from $20 to $200 per month, removing the need to maintain a desktop setup or VPS. That is a sensible on-ramp for users who want the benefit of autonomous workflows without becoming amateur systems administrators. It also gives the company recurring revenue and a direct relationship with customers, rather than relying entirely on goodwill from an open-source community.
There is an unavoidable tension, though. The more valuable Hermes becomes, the more users will ask for easy setup, polished integrations, enterprise controls, and dependable support. Those features generally favor a hosted product. Yet the people drawn to open agents often want local control precisely because they do not want a company sitting between their work and their AI. A larger Nous Research funding round would give the company more resources to manage that balance, but it would not make the tension disappear. Nous will need to avoid turning its most enthusiastic users into spectators.
Its approach is visible in the broader project ecosystem: the Nous Research GitHub organization includes models and tooling intended for developers who want more than a consumer-facing assistant. That technical credibility buys Nous some room. The commercial question is whether the company can package it without sanding off the openness that made it popular.
What investors are really buying
Nous Research funding is not only a wager on Hermes. It is a wager that the agent layer becomes as important as the model layer. The biggest labs have enormous advantages in training frontier models, distribution, and compute contracts. Smaller companies need to win somewhere else: better tools, more flexible deployment, sharper developer communities, or a product that people use every day because it actually gets work done.
My read is that Hermes has landed on a plausible opening. It sits where today’s AI enthusiasm is hottest: coding, personal automation, and remote workflows. But agent startups now face a brutal test. Demos are easy. Reliable systems that survive messy real-world permissions, changing websites, vague instructions, and adversarial inputs are much harder.
If the round closes at the reported $1.5 billion, Nous will have the resources to chase that test aggressively. The valuation also means the company has less room for a merely decent outcome. That is the pressure created by Nous Research funding at this scale. Hermes needs to become a tool people trust with consequential work, not just a clever agent people show their friends in Discord. That distinction will decide whether this is smart financing or another reminder that AI investors can get ahead of the product.
