Private by default
Your mail, your calendar, your tasks, your money: watched by an assistant that works for you and reports to no one. This is the short story of how.
What it does
No feeds, no setup maze, no forms. You open Anjadhe and there is one thing on the page: a place to say what you need.
Tasks, dates, and the goal behind them all land in your schedule, where you and the assistant manage them together.
Connect Gmail and the assistant quietly reads what arrives, then files what matters into folders you can trust:
Deadlines become dated tasks automatically. You never miss an important email because you never had to catch it.
The assistant interviews you about goals and risk, writes the strategy down, and then the math does the watching:
Adherence is computed, not judged by the model. A morning routine reviews the market in the context of what you actually hold, and flags a trade that would take you off plan.
A routine is a standing instruction in plain words. Say what you want and when, and Anjadhe keeps doing it without being asked again. The results post to your Home feed:
You approve each routine once when it is created, and you can pause or change any of them on the Routines page. These posts are real runs over a demo week.
Journaling, notes, bookmarks, a health log, calendar, news briefings, and routines that run on their own: morning digests, weekly goal reviews, an eye on the inbox.
Everything the assistant can touch, you can also do by hand. It is a coworker, not a chatbot.
How it’s built
No chat log to scroll back through. The data the assistant manages appears as minimal, modern apps: a task list that looks like a task list, a portfolio that looks like a portfolio.
A shared canvas: the assistant writes into the same records you edit, and both of you see the same truth.
We never see your records, and we have nothing to breach, subpoena, or monetize. It is also how the core product can be free: there is no per-user database to pay for.
A claim you can test: pull the network cable, and the apps, your data, and a local brain all keep working.
No Anjadhe account, no email verification, no password. Download it, open it, start.
An assistant that can do real things needs rules about doing them. Anjadhe’s are built into the app, not written in a policy:
You can grant standing permission for steps you trust. That is your call to make, and it is never the starting point.
A large language model is the assistant’s brain, and capable ones are big. So Anjadhe gives you three honest options:
An open-weight model running locally. Nothing leaves the machine. Works best with 32 GB of RAM or more.
A managed open-weight model for Macs with less memory. A free monthly allowance, no account needed. Paid capacity may come later.
Plug in your own Anthropic or OpenAI key for a frontier model. Always an explicit opt-in, never a default.
There is no silent fallback: the model you chose is the one that answers, and switching is always your action.
Your chats, and the local data the assistant reads to answer you, are sent to a remotely hosted model. That is the trade. Two things make it a fair one:
Prompts and answers are never written to our logs or stored on our servers. The service exists to relay, not to remember, and its source code is public so you can check exactly that.
The model providers we use run zero-retention inference: what the assistant sends is not stored and not used to train models.
Private by default, and honest about the rest. Your data goes only where you point it.
The deal
We don’t collect your personal data and we don’t sell it to anyone. The source code for the app and for Anjadhe Cloud is public on GitHub, so you don’t have to take our word for any of this.
Our goal is to build a personal AI assistant you can trust completely, so you can let it all the way into your life and get the full benefit of what AI can do.
Anjadhe · a personal AI assistant, private by default.
Anjadhe