Autonomie is a self-hosted system that watches how you interact with your computer and slowly learns to act on your behalf.
It doesn't ask you to write scripts or hardcode macros.
It watches, remembers, and evolves based on how you use your machine.
Autonomie's goal is simple: your PC should already know what you were going to do.
This is not traditional automation. This is the beginning of a digital self — your second set of eyes, hands, and intent living inside your OS.
Today, all advanced user modeling is controlled by corporations:
- They predict what you’ll click — but only to sell you something.
- They record your behavior — but never share what they learn.
- They build personal AI — but it's not yours.
You’re being studied by your machine, yet you have no access to the brain behind the glass.
You deserve the same power — on your terms.
Autonomie empowers you with a system that:
- Observes your real behaviors (no manual programming)
- Learns visually from what you see
- Remembers your interactions
- Evolves your preferences
- Lives locally and only learns from you
This is about freedom, agency, and control in your digital life.
Autonomie is a step toward building the first true digital human interface — a machine that mimics your real interactions in every sense.
It’s not just automation. It’s a replication of how and why you interact.
| Human Trait | Autonomie Counterpart |
|---|---|
| Vision | Screenshot analysis + object detection + OCR |
| Memory | Visual memory graph with element IDs and frequency |
| Touch / Movement | Mouse path simulation, wheel scrolling, timing variance |
| Attention | Top-down screen parsing and UI relevance heuristics |
| Decision Making | Action prediction based on past behavior (in development) |
| Emotion/Bias | Weighted action preference by time of day, hesitation, etc |
The computer shouldn't just run tasks. It should run your intent.
Avoiding CAPTCHA is not the goal, but it is a necessary obstacle for truly human interfaces.
Autonomie avoids API-level "captcha solvers" because they:
- Compromise your data
- Violate terms of service
- Use third-party behavioral data
Instead, Autonomie achieves authenticity by actually behaving like a human:
- Top-down visual scanning
- Scroll wheel behavior with irregular timing
- Mouse movement that includes acceleration, wobble, and pauses
- Click hesitation and overshoot correction
- Response to visible elements only, just like human vision
- Live OCR to mimic reading text, not querying metadata
If all else fails, Autonomie uses vision + interaction delay to legitimately pass human checks — not because it's faking, but because it's behaving naturally enough to be treated like you.
If a CAPTCHA can’t tell the difference, maybe there isn’t one.
- UI Object Detection – Detect buttons, labels, checkboxes, and more using OpenCV.
- Text Recognition (OCR) – Read visible screen text with Tesseract.
- Visual Memory System – Recognize elements across sessions with hashed visual IDs.
- Live CLI – Pause/resume/inspect engine state in real-time.
- Behavior Logging – Record user behavior for pattern analysis (non-invasive).
- Scroll Logic – Dynamic wheel behavior with noise functions.
- Mouse Movement Simulation – Realistic pathing and speed variability.
- Intent Modeling – Learn patterns like “what I usually do after I open this app.”
- Task Recognition – Notice and offer to automate common behavior chains.
- Autonomous Response – Let the system take action when confidence is high.
- Neural Agent Mode – Use lightweight LLMs to ask: “What would you do here?”
git clone https://github.com/unaveragetech/autonomie.git
cd autonomieThis installs required Python packages and prepares the local environment:
bash setup.sh- Windows: Add install path to Environment Variables.
- Linux/macOS:
export PATH=$PATH:/path/to/tesseractStart the main interface:
python main.py| Command | Description |
|---|---|
exit |
Safely shuts down the engine |
pause |
Temporarily stops activity without quitting |
resume |
Continues activity after a pause |
status |
Prints current engine state |
toggle_logging |
Enable/disable console action logs |
clear_logs |
Clears the CLI output for easier reading |
help |
Lists all available commands |
This project is early-stage but ambitious. We’re building a prototype of the future — and we want you to help shape it.
Ways to contribute:
- ⭐ Star the repo to support visibility
- 🐛 Report bugs or feature requests
- 🔧 Submit PRs with improvements or integrations
- 📣 Share the vision with others in open-source, privacy, and AI communities
Want to help design the UI layer? Or plug in your own vision model? Or build the first plugin system?
This is your invitation. Let’s build you — in code.
This project is licensed under the SDUC License (Self-Directed Use & Contribution). See LICENSE for terms.
Autonomie is not a tool. It's a statement:
You should own the system that learns you.
If corporations build AI to predict your every move, you should have that power too — on your desktop, with your data, and under your control.
Let’s build the digital self. One screen, one mouse movement, one learned habit at a time.