Everything you need
to run InvisiQ.
A practical guide to installing InvisiQ, signing in, starting your trial, connecting your own AI key, and driving the overlay from the keyboard. Beta build v1.2.0, Windows 10 v2004+ and Windows 11.
From download to overlay in under a minute.
InvisiQ is a single portable .exe. No installer, no admin rights, no registry entries. Here is the exact order you will see on first launch.
Download and run
Grab the portable build from the download page and double-click it. Nothing installs to your system.
Continue with Google
Sign-in opens your system browser, so there is nothing to type into the app. You stay signed in on later launches. This starts your 14-day free trial.
Accept "Before you start"
A one-time beta-data notice explains what is stored during the beta. You accept it once. Full detail lives on the Privacy page.
Onboard, then summon
A 3-step wizard helps you add an API key, see the hotkeys, and run a stealth self-test. Then press Ctrl ⇧ G from any app.
Free for 14 days. No card.
Bring your own key, paste it once.
InvisiQ never ships an AI key of its own. Generate one from a provider you trust, paste it into Settings → Providers, and it stays encrypted on your device.
Keys are encrypted on your device with AES-256-GCM (a machine-derived key, PBKDF2-SHA256, 600K iterations). They never leave your machine except as the body of a direct call to the provider you chose. Remove a key in one click anytime.
Seven models, switchable mid-chat.
Pick a model and provider from the header dropdown at any time. Vision models read screenshots directly; o3-mini is text and reasoning only.
| Provider | Model | Vision | Context | Speed |
|---|---|---|---|---|
| OpenAI | GPT-4o | yes | 128K | medium |
| OpenAI | GPT-4o Mini | yes | 128K | fast |
| OpenAI | o3-mini | no | 200K | slow |
| Anthropic | Claude Sonnet 4 | yes | 200K | medium |
| Anthropic | Claude Haiku 4.5 | yes | 200K | fast |
| Gemini 2.0 Flash | yes | 1M | fast | |
| Gemini 2.5 Pro | yes | 1M | medium |
InvisiQ is driven from the keyboard.
These are the defaults. Every one is editable in Settings → Hotkeys, so you can rebind anything that clashes with the apps you already use.
| Ctrl ⇧ G | Toggle overlay visibility |
| Ctrl ⇧ S | Capture full screen and send to AI |
| Ctrl ⇧ R | Capture a dragged region and send to AI |
| Ctrl ⇧ A | Focus the text input |
| Ctrl ⇧ C | Copy the last AI response |
| Ctrl ⇧ V | Paste the last response into the active app |
| Ctrl ⇧ N | New conversation |
| Ctrl ⇧ I | Toggle stealth typing / capture mode |
| Ctrl ⇧ P | Toggle click-through (mouse passes through) |
| Ctrl ⇧ ] / [ | Next / previous model |
| Ctrl ⇧ Q | Panic: exit capture, remove the hook, hide everything |
| Esc | Hide the overlay immediately |
One-tap personas for the task at hand.
Each mode tunes the AI's behavior with a system prompt and sensible defaults. You can also define your own custom modes.
General
The everyday default. Light prompt, fast responses, your choice of voice.
Coding
Code-aware prompt, syntax highlighting, and paste-to-editor for snippets.
Meeting
Notetaking shortcuts, agenda recall, and action-item extraction.
Solve
Step-by-step problem solving: approach, worked solution, and reasoning.
What is in the beta build.
Invisible overlay
A floating panel excluded from screen capture at the OS level. It shows on your monitor but is stripped from any shared screen, recording, or screenshot.
Ask about your screen
Capture the full screen, a dragged region, or an in-overlay snip. The image goes to a vision model and the answer streams back.
Streaming chat
Token-by-token answers with Markdown, syntax-highlighted code, one-click copy, and preserved multi-turn context.
Stealth typing
Type into the overlay from any app without it stealing focus, so the panel never becomes the foreground window a monitor would notice.
Click-through & panic
Let the mouse pass through the overlay while you work, hit Escape to hide instantly, or use the panic hotkey to exit capture and hide everything.
Memory (RAG)
A built-in TF-IDF memory auto-extracts facts and feeds relevant context back into prompts, encrypted on your device.
- Conversation history with search and export (JSON, Markdown, TXT, PDF)
- Smart paste straight into the active app
- Voice and meeting assist: speech-to-text, live transcript, system-audio capture
- 20+ prompt templates across 8 categories, plus your own
- Companion mode: drive the overlay from your phone over a local link
- Multi-monitor with hot-plug detection
- Dark and light themes, with a per-request and per-session cost meter
- Auto-update so beta builds stay current
A documented Windows privacy feature.
Windows draws every window through a compositor. InvisiQ asks Windows to put its overlay on a layer that is sent to your monitor but withheld from anything that captures the screen. The pixels reach your eyes, not the screen-share, the recorder, or the screenshot. It is a documented privacy feature, not a hack or a driver exploit. Read the full breakdown on the Security page.
Honest limits. InvisiQ is invisible to standard screen capture, sharing, and recording APIs and to foreground-window monitoring on Windows 10 v2004+. It is not magic: a physical camera pointed at your monitor will still see the overlay, and kernel-level enterprise DLP agents can still enumerate running processes. We never claim it is "undetectable" or "guaranteed invisible everywhere."
If you can run Windows 11, you can run InvisiQ.
Why you may see "unknown publisher".
The beta is an unsigned build, so on first run Windows SmartScreen may show "Windows protected your PC, unknown publisher." This is expected for an unsigned beta and does not mean the app is unsafe. Code signing arrives once a company entity exists.
You do not need to disable Windows Defender or add any exclusions. Just the two clicks above.
What is stored, and what never is.
Your screen never leaves your device
Screenshots and screen contents are never stored or uploaded to InvisiQ. Whether a prompt included an image is recorded only as a yes/no flag; the image itself is not kept.
Prompt text during the beta
The text of the prompts you send is stored on our backend to help us improve the product. API keys and obvious personal info are stripped first, it is deleted after 30 days, and you can wipe it anytime in Settings → Privacy.
Local-first by default
Your API keys, conversations, and memory index stay encrypted on your machine. AI calls go directly to the provider on your own key. We also collect privacy-safe usage analytics.
This is the same notice you accept in the app: "During the beta, InvisiQ stores the text of the prompts you send, so we can understand what to build and improve the product. We store text only, never your screenshots or screen contents. API keys and obvious personal info are stripped before storage. Prompt data is deleted after 30 days, and you can wipe yours anytime in Settings → Privacy." Read the full Privacy Policy.
Common questions.
Still stuck? Email hello.invisiq@gmail.com.
A few terms, defined.
Read the answer.
Skip the awkward pause.
Download InvisiQ, sign in with Google, and start your 14-day free trial. No card required.