Developer builds "Anna," a Windows AI assistant with Telegram and email, on top of OpenClaw

A developer has shown a prototype of "Anna," an AI assistant for Windows meant to run constantly on a home computer and help the user not only in chat but also in the background. The project is conceived as a friendlier alternative to complex agent systems: instead of a terminal and manual configuration, it is an ordinary desktop app with a clear setup wizard.

From idea to product

Two things prompted the project: the film "Her" and the author's interest in OpenClaw. Extra momentum came from the buzz around OpenClaw, which by then had gained about 300 thousand stars on GitHub. After getting to know this stack, the idea of a "home AI operating system" no longer looked like pure science fiction: an assistant can already converse almost like a person and carry out useful actions.

But the limitations of this approach showed up very quickly. To get started, you need to understand the internal mechanics of nodes, channels and the links between components, and setup runs through a text wizard in the terminal. For an engineer that is tolerable; for a mainstream user it is almost a guaranteed barrier.

In response, the author decided to build not yet another set of scripts but a product for people with no technical training. The logic is simple: the user installs the program on Windows, goes through step-by-step onboarding, connects the services they need and describes in a few words what they want from the assistant. After that, "Anna" is supposed to live alongside the user permanently, accept commands from the app window or Telegram, and reach out on her own when something important happens, for example when an email arrives, a product's price changes or a reminder comes due.

What is already done

At this stage the project has a working prototype in the form of a full desktop app built on Electron and React. It is not a console build for enthusiasts but a familiar Windows program with a graphical interface, a built-in chat and local storage of conversation history. The dialogue is not lost after a restart, and initial setup is handled by a separate startup wizard. In other words, the user does not need to hunt for configs by hand, edit environment variables or work out which service connects to what.

• A full Windows app with a GUI
• Chat with locally saved history
• Creation of background tasks in plain language
• Integration with Telegram and email

The second important part of the MVP is the shift from an ordinary chatbot to a persistent assistant. The prototype can already take natural phrasing such as "remind me in an hour" or "check this every day" and turn it into background tasks. The project also has external communication channels: Telegram for everyday conversation and email for reading, displaying and drafting messages. As a result, "Anna" works not as a separate window on the desktop but as a service that can watch for events and return to the user only at the right moment.

The next MVP stage

Next, the author wants to strengthen practical usefulness rather than demo effects. The roadmap centers on scenarios an ordinary person can really use every day: reminders, email triage, price tracking, search in personal files and a morning summary of the day. To that end, the project plans to develop the step-by-step onboarding and add connections to basic data sources: weather, news, Brave Search and Google Calendar. The assistant would then be able to assemble, on request, not an answer "from memory" but a short, current picture of the day from real services.

"I don't want to build a project around magic for magic's sake."

A separate priority is RAG search over local documents and smarter background monitoring. The idea is that a user can ask about an old contract, a note or a manual without copying text into the chat by hand, and the system finds the right fragment in their folders on its own. In parallel, the author wants to teach "Anna" to track prices, product availability, website changes and important emails, and eventually to launch Chrome, open ordinary Windows applications and perform permitted actions on the home PC. That moves the project closer to a personal agent than to a chat interface.

What this means

The "Anna" project is a good illustration of where the market for personal AI tools may shift: from impressive but complex agent frameworks toward local assistants with simple onboarding, understandable integrations and proactive behavior. The main question is not how impressively a model answers, but whether a person without a technical background will be able to use such a helper every day. If so, the next step for AI products is not yet another chat but a permanent digital layer on top of the computer and personal services.