How a Habr author built a Hermes Agent assistant in Telegram for his family and customer support

A Habr author shared his personal experience using the Hermes Agent AI agent in Telegram, from a side project he runs with his wife to other tasks, including customer support.

How the project started

According to the author, the idea grew out of a small joint project with his wife: she handles the creative side, and he handles the code. After building the project's scaffold in Codex, the author set up a three-way group chat with himself, his wife, and Hermes Agent, then gave the agent access to the project and its context. As a result, his wife can now add features and move interface elements on her own by sending voice messages straight to the shared chat.

• The project scaffold was built in Codex
• A three-way Telegram group chat was created: the author, his wife, and Hermes Agent
• The wife makes interface changes through voice messages without writing a single line of code
• According to the author, the experiment worked so well that he expanded his use of Hermes Agent to other scenarios

Where else Hermes Agent proved useful

The author notes that the success of the family project pushed him to use Hermes Agent in other scenarios as well, including customer support and personal side projects. In the article, he promises to explain what Hermes Agent is, how exactly he uses it, and how readers can set up a similar assistant for themselves.

Why this matters to developers

The key detail of this case is that the author-developer isn't the only one with access to the agent: so does someone with no technical background. The author's wife talks to Hermes Agent by voice rather than through code or a console, and the agent independently turns her spoken instructions into concrete changes to the project's interface and functionality. The author considers this kind of collaboration — a developer and a non-technical person working through the same agent — the most valuable result of the experiment.

Crucially, the whole process runs without intermediaries relaying requests. The wife doesn't have to explain to her husband what needs fixing in the interface, and the husband doesn't have to turn her requests into a task for himself. Both talk directly to the same agent, which has shared access to the project's code and context, and the agent carries each change through to completion on its own.

What's behind the "group chat with an agent" format

The format the author describes isn't a separate app or a bot with a fixed set of commands. The agent is a full participant in a shared Telegram chat, on equal footing with the humans. Both people see the same conversation with the agent, either of them can step into a task at any moment, and the agent responds to messages from both participants within the shared project context. This differs from the classic setup, where each developer has separate access to an assistant and a separate conversation with it. Here, the entire context is shared and synchronized across the two-person team.

What it means

The case shows how AI agents like Hermes Agent are lowering the barrier to entry for software development: a person with no technical background can shape a product directly by voice, without writing code, as long as the agent has access to the right context and tools. For small teams and family side projects, this means the role of "second developer" can be filled not by a person but by an agent that both participants trust with the same set of permissions. The author stresses that this effect — rather than any single feature or line of code — was the main discovery of the entire experiment.