Nous Research Launches 'Hermes Agent' Open-Source Framework

Nous Research, the team behind Hermes Agent, is an AI research lab with its headquarters in New York City, and they are actively hiring for roles like Research Scientist. Founded in 2023, the company is venture-backed, having raised $70 million in funding from investors including Paradigm, and is focused on developing open-source, human-centric AI models. The Hermes Agent framework is engineered for persistence, moving beyond the limitations of single-session chatbots. It utilizes a multi-level memory architecture that allows it to retain information across sessions and even automatically generate "Skill Documents" when it successfully completes a complex task, effectively creating a reusable knowledge base. This is a key differentiator from many other frameworks that have limited or session-based memory. For developers looking to build, Hermes Agent provides a robust sandboxing environment with five different backends, including Docker, SSH, and local execution, allowing for safe and flexible development workflows. It comes with over 40 built-in tools for web browsing, code execution, and task planning, and its functionality can be extended through a skills ecosystem that uses the open standard `agentskills.io`. The framework's GitHub repository includes documentation and examples for getting started, including instructions for adding custom functions. Engineers bootstrapping side projects with AI agents often emphasize starting with a very specific, annoying task you want to automate, rather than trying to build a general-purpose assistant. This approach of "atomization," breaking down a problem into small, deterministic steps, leads to more reliable results than large, all-encompassing prompts. For those working a full-time job, time-boxing AI experiments and using automation tools for repetitive tasks are crucial for maintaining focus and avoiding burnout. The NYC startup scene offers several resources for aspiring founders in the AI space. Accelerators like AIR are specifically looking for design-obsessed founders building AI products and offer investment and a 10-week program in the city. For those with a more established idea, the NYC AI Landing Pad is a 12-month program for global founders ready to scale. Additionally, New York State's NY Ventures offers a Pre-Seed and Seed Matching Fund Program for early-stage tech companies. When it comes to building a vertical SaaS or consumer app with AI agents, the key is to focus on a niche workflow and gather domain-specific data to train your agent. For vertical SaaS, this could mean automating tasks in finance or healthcare. For consumer apps, developers have found success by building AI companions that handle specific user needs, but they caution that managing hallucinations and ensuring a reliable user experience requires careful, iterative development. For those considering the leap from enterprise to a startup, the experience of being a founding engineer at an AI startup involves a product-first mindset and a willingness to take on work outside of a narrow job description. Indie hackers have successfully built and launched AI agent-based businesses, with some reaching monthly recurring revenue by creating "AI agent factories" that allow non-technical users to build their own agents. These stories highlight the potential for engineers to turn side projects into viable businesses. When pitching an AI agent startup to VCs, especially in the competitive NYC market, investors are looking for more than just a GPT wrapper; they want to see a unique value proposition, proof of customer traction, and a clear "moat" or defensibility. Some VCs are now using their own AI agents to analyze pitch decks and founder updates, indicating that a deep understanding of AI is becoming a prerequisite for both building and fundraising.