Most AI chat tools forget you the moment a conversation ends, and many keep your prompts on servers you know little about. Over the past weeks we have been testing an AI agent that works differently on both counts. It is called Hermes, it is now in beta testing on the bacher-ai platform, and it was set up with privacy as the starting point rather than an afterthought.
What Hermes is
Hermes Agent is an open-source AI agent from Nous Research. Unlike a plain chatbot, it does not just answer and forget. It builds a memory of the people it works with and the projects they care about, so you do not have to explain your context again in every new chat. It can search the web, keep to-do lists, work with documents, look at images and learn reusable skills for recurring tasks. It can also run scheduled jobs, such as a short daily summary delivered every morning.
On the bacher-ai platform, Hermes appears as a regular model in the chat interface you already know. You pick it from the model list and talk to it like any other assistant. The difference is that it remembers, uses tools and works towards a result instead of producing a single reply.
Privacy built in from the start
An agent that remembers you also knows a lot about you, so where your data goes matters even more. Our setup follows a few simple rules. In the platform chat, Hermes runs on models inside a trusted execution environment: a sealed hardware enclave designed so that not even the provider’s operators can look at what is being processed. Other requests go through a European endpoint with zero data retention, which means the model providers do not store your prompts or use them for training.
Hermes also has its own memory and persona, kept separate from the rest of the platform, and its memory stays on our own infrastructure. The agent runs in an isolated environment with only the tools it needs, and every model key has a strict spending limit. We are also testing a mode in which Hermes asks for your approval before it uses a tool.
Beta now, platform soon
Right now Hermes is available to a small group of testers. We are using this phase to see how well its memory works in daily use, which tools are actually helpful and where the guardrails need adjusting. Once we are confident it is useful and safe, Hermes will become available to all users of the bacher-ai platform. We will announce it here when that happens.
AI agents are becoming more capable quickly, and with that comes more access to personal information. Hermes is our attempt to show that a helpful, personal agent and strong privacy can go together.
Key Takeaways
- Hermes is an open-source AI agent with memory, tools and skills, now in beta on the bacher-ai platform.
- It appears as a regular model in the familiar chat interface.
- Platform chats run on models inside sealed hardware enclaves; other requests use a European endpoint with zero data retention.
- Its memory stays on our own infrastructure, separate from the rest of the platform.
- After the beta, Hermes will be available to all platform users.
Photo: Alex Knight via Pexels
