Chronos is a voice-powered AI assistant for Windows that launches your applications on command — no cloud, no tracking, no compromise. This document defines exactly how your privacy is protected.
Chronos is a locally-executed voice assistant built for Windows 10 and 11. It listens to your spoken commands, interprets them using the Vosk speech recognition engine, and launches or controls your applications — all without sending a single byte to external servers.
Powered by Vosk, an offline-first speech recognition engine. Vosk models are downloaded once and run entirely on your CPU. No microphone data ever leaves your device.
Chronos responds using KOKORO TTS — a neural voice that closely mirrors the iconic Jarvis tone. The voice synthesis runs within the Edge runtime already on your Windows system.
Chronos executes local system commands to open, switch, and manage applications installed on your PC. All command resolution happens on-device via an editable local app registry.
At first launch, Chronos may fetch the Edge-tts voices (one-time download). After this setup, the system transitions to fully offline operation — no recurring network traffic.
After the initial model download, Chronos has zero cloud dependencies. There are no API calls, no telemetry pings, no usage analytics, and no crash reporting sent to any server.
Your command history, app list, and preferences are stored exclusively in local configuration files on your machine. You can view, edit, or delete them at any time.
Understanding the data lifecycle within Chronos — from the moment you speak to the moment your app launches.
Your system microphone captures audio. This audio stream is piped directly into Chronos's local processing pipeline in memory. The raw audio is never written to disk, uploaded to any server, or retained after the recognition process completes.
RAM only · never persistedThe audio is transcribed using the Vosk engine running entirely on your CPU. Vosk is an open-source, offline-capable speech recognition toolkit developed by Alpha Cephei. The language model (typically 40–500 MB depending on selected model) lives in your local AppData folder and performs inference without internet access.
Offline · open source · CPU-boundThe transcribed text string is matched against your local command registry — a configuration file listing your apps, aliases, and custom shortcuts. This matching is purely string-based pattern logic, executed on-device with no external model calls.
Local config · editable by userChronos calls the Windows API to launch or interact with the target application. It uses standard subprocess invocation — the same mechanism as double-clicking an icon. No elevated privileges are required or requested beyond what the application itself needs.
Windows API · no elevation requiredChronos confirms the action using Microsoft Edge's built-in TTS engine. This voice synthesis is processed using the neural TTS runtime bundled with Microsoft Edge — the same voice engine that powers Edge's read-aloud feature. For offline use, Chronos falls back to Windows SAPI voices available on your system.
Edge runtime · local SAPI fallbackA complete, transparent breakdown of every category of information Chronos interacts with and where that data goes.
| Data Type | Purpose | Storage Location | Retention | External Access |
|---|---|---|---|---|
| Microphone audio stream | Speech recognition input | RAM (transient) | Discarded after recognition | NEVER |
| Transcribed command text | Intent matching | RAM (transient) | Discarded after execution | NEVER |
| App registry / command config | Command resolution | %AppData%\Chronos\config.json | Until user deletes | LOCAL ONLY |
| User preferences & settings | Customization | %AppData%\Chronos\settings.json | Until user deletes | LOCAL ONLY |
| Vosk language model | Offline speech recognition | %AppData%\Chronos\models\ | Persistent (user-removable) | DOWNLOADED ONCE |
| Command execution logs | Debugging (opt-in only) | %AppData%\Chronos\logs\ | 30-day rolling (if enabled) | LOCAL ONLY |
| Crash reports | Not collected | N/A | N/A | NEVER |
| Usage telemetry | Not collected | N/A | N/A | NEVER |
| Personal identifiers | Not collected | N/A | N/A | NEVER |
Privacy is only as strong as the security underneath it. These are the technical safeguards built into every release of Chronos.
Chronos is built on trusted, auditable open-source technology. No black-box SDKs. No proprietary analytics libraries.
Developed by Alpha Cephei. Vosk is open-source (Apache 2.0) and runs 100% on-device. The model is downloaded once from the official Vosk model repository and stored locally. No Vosk servers are contacted during recognition.
Chronos uses the edge-tts Python library which interfaces with the Microsoft Edge Read Aloud feature. This library sends the response text to Microsoft's TTS service during synthesis. For fully offline operation, Chronos falls back to Windows SAPI voices installed on your system.
Because all your data stays on your machine, you already have the highest level of control possible. These are the specific actions you can take.
Navigate to %AppData%\Chronos\ in File Explorer at any time to view all configuration files, logs, and model data stored by Chronos. No special tools required — it's all plaintext JSON and binary model files.
Your app registry, voice preferences, wake-word settings, and all other configuration values live in human-readable JSON files. Open them in any text editor (Notepad, VS Code) and modify them freely. Chronos reloads on next launch.
Uninstalling Chronos or manually deleting the %AppData%\Chronos\ folder removes everything completely. There is no server-side data to delete because none was ever sent.
Command logging is disabled by default. If you enabled it for debugging, set "logging": false in settings.json or toggle it from the Chronos system tray menu. Existing logs can be deleted from the logs subfolder.
After the initial Vosk model download, enable full offline mode in settings. In this mode, Chronos uses Windows SAPI for voice output instead of Edge TTS, eliminating all network activity. The assistant functions identically — only voice quality differs.
We encourage you to verify Chronos's network behavior using Windows Firewall logs, Wireshark, or GlassWire. With offline mode enabled, you should see zero external connections from the Chronos process. The Vosk model download is the only expected outbound request.
Chronos does not collect any personal information from anyone, including children under the age of 13. Because Chronos operates entirely locally and collects no user data whatsoever, COPPA compliance is inherently satisfied. No registration, no account, no data — for users of any age.
If this privacy policy changes, the updated version will be posted with a new effective date. Since Chronos does not collect contact information, we cannot notify you directly — we recommend bookmarking this page if you wish to stay informed. Breaking changes in data handling will always accompany a major version update with release notes.
Because Chronos does not transmit or store data on external servers, no cross-border data transfers occur. This policy is published in English and is intended to meet the spirit of major privacy frameworks including GDPR and the Worldwide Consumer Privacy rules, even though traditional compliance obligations do not apply to a fully local application.
We believe trust must be earned through transparency. The Chronos source code is available for review on GitHub. You are encouraged to audit the codebase, inspect network calls, or run it through any static analysis tool. If you identify a discrepancy between the code and this policy, please open an issue immediately.
If you have any questions about how Chronos handles your data, want to report a privacy concern, or would like clarification on any point in this document, reach out directly. We respond to all privacy-related queries within 72 hours.