Navisual
Your step-by-step guidance for any task, in any app.
Ask a chatbot how to do something — and you're still hunting for the button it means. Let an AI agent do it for you — and you give up control of your computer. Navisual does neither: it understands what you see on the screen, works out what to do, and shows you where — one step at a time.
Download for Windows 30 free requests · No sign-up · Signed installerUser guide · View on GitHub · Release notes
The AI guides, you decide.
Navisual uses AI, which can make mistakes — always verify each suggested action before performing it.
Works for any task, in any app
Follow instructions you already have
"Here are the setup steps from the project README — walk me through them."
Paste a support article, a runbook, or a chatbot's long answer. Navisual turns the wall of text into one step at a time and follows you from app to app — browser, editor, terminal — putting each command on your clipboard so you paste instead of retype.
Office & productivity
"Add a pivot table to this spreadsheet."
Walks you through Insert → PivotTable, field setup, and formatting step by step. Excel, Word and PowerPoint get extra-precise pointing.
Learning new software
"How do I add a bevel to this cube in Blender?"
Points at the right tool and narrates the shortcut — learn by doing, not by reading a manual. An optional Blender add-on makes the pointing exact.
Everyday computer help
"Help me set up my email account."
Guides through account wizards, settings pages, and unfamiliar dialogs — and answers in your own language, whatever language the app is in.
Navisual is designed to work with any Windows application. Some apps are more deeply optimized than others — more optimization is on the way.
How it works
Type what you want to do — or say it out loud with push-to-talk.
Navisual takes a quick look at your active window, asks the AI for the next step, and points at the exact button to click.
Follow the pointer and audio narration. Press Ctrl+` when done to advance to the next step.
Works with your preferred AI
30 free requests included. Bring your own key for unlimited use — or run fully offline with Ollama.
Privacy
Navisual sends only your active window — never your full desktop — and only to the AI provider you choose. Screenshots stay in memory and are never written to disk.
The free tier is free because its AI provider may retain and train on your requests; paid tiers, your own API key, and offline Ollama don't. Full details in our Privacy Policy.
Frequently asked
What does Navisual do, and who is it for?
It watches your active window and gives you real-time, step-by-step guidance for whatever task you describe — pointing at the exact button or menu to click next, with audio narration. It's for anyone stuck on a task in an unfamiliar app: learning new software, walking a family member through a setup wizard, or just not remembering where a setting lives. Think of it as someone experienced sitting next to you and pointing — except you're always the one clicking.
Does the AI control my mouse?
No. Navisual only points at where to click and narrates the next step. Every click, keystroke, and confirmation is yours.
Does it work in the browser, and with the command line?
Both. Web pages are among the surfaces Navisual reads most reliably — web apps, forms, admin panels and settings pages all work, and Chrome and Edge get a dedicated fast path. For the terminal, Navisual never runs anything: when a step involves a command, it puts the exact text on your clipboard so you can read it, then paste and press Enter yourself. Same for URLs, filenames, and anything else you'd otherwise retype. A task that spans a browser download, a text editor and a terminal is fine — Navisual follows whichever window you're working in.
What data leaves my machine?
A screenshot of your active window (not the full desktop) plus your conversation text, sent to whichever AI provider you choose. Element matching runs locally. Note: the free tier's AI models may retain and train on those requests (screenshot included); paid tiers don't under their providers' current policies, and Ollama keeps everything local — see the Privacy Policy.
Can it run fully offline?
Yes. Choose Ollama as your provider and pull a vision model (e.g. llama3.2-vision). Nothing leaves your machine.
Which Windows versions are supported?
Windows 10 (1809 or later) and Windows 11. WebView2 is required — it ships with Windows 11 and installs automatically on Windows 10.
Is it open source?
Source-available under FSL-1.1-Apache-2.0. Each version converts to Apache 2.0 two years after release.
Windows says “Windows protected your PC” — is it safe?
Yes. That’s Microsoft SmartScreen, which warns on any newly-released app until it has been downloaded enough to build reputation — it’s about download volume, not safety, and it appears even for properly code-signed apps. Navisual’s installer is code-signed (Azure Trusted Signing) and fully source-available (FSL). To continue, click More info, then Run anyway. The warning disappears as more people install.
What happens after the free 30 requests?
Top up the managed tier with pay-as-you-go coins (from $5), bring your own API key (Anthropic, Gemini, OpenAI, DeepSeek, Qwen), or run locally with Ollama for unlimited free use.
macOS or Linux?
Windows only for now. macOS and Linux ports are planned after the 1.0 public launch.
支持中文 / DeepSeek / Qwen 吗？
支持。在设置中选择 DeepSeek 或 Qwen，填入 API Key 即可。DeepSeek 仅传输文字（不上传截图）；Qwen 支持截图理解。