Use case: coding and pairing

An AI coding assistant overlay for Windows

AuraCloak floats an AI chat above your IDE, terminal and browser. Press Alt+S to screenshot the stack trace, ask GPT, Claude or GitHub Copilot what broke, and read the answer in syntax-highlighted code blocks while the window you are working in stays put.

Screenshot any windowGitHub Copilot sign-inLocal models via Ollama
Beyond the editor

Your IDE assistant only sees your IDE

Copilot in VS Code knows your open files. It can't see the failing CI log in the browser, the Grafana panel on your second monitor, the 400 response in Postman, the PDF spec, or the legacy app running in a remote desktop window. A lot of debugging happens in those places.

An overlay solves this with one hotkey. Alt+S hides AuraCloak for a quarter of a second, captures every monitor and attaches the image. Alt+Shift+2 grabs just monitor two. Type the question and the answer streams back with code blocks you can copy.

It also helps when people are watching. Share your IDE in a pairing session, a code review or an OBS stream, and your questions to the AI stay on your monitor, because Windows leaves the overlay out of the capture.

Scenarios

Where developers use an AI coding overlay

Stack traces in the terminal

Screenshot the traceback and ask for the likely cause and the smallest fix. Faster than selecting 80 lines in Windows Terminal.

CI logs and dashboards

Point it at the red build log or the latency graph in the browser and ask which step or service changed.

Pairing over screen share

Check an API signature or a flag you forgot while your partner watches your editor, without opening a browser tab on the shared screen.

Streams and tutorials

OBS skips the overlay, so you can ask for help mid-stream and decide what to show.

Proprietary code

Point AuraCloak at Ollama or LM Studio on localhost. Prompts and screenshots stay on your machine.

Code you can't copy

Remote desktop sessions, screenshots in tickets, code in a video. A screenshot gets it to a vision model.

Example prompts

Typical prompts while coding

Replies are examples and will vary by model.

Traceback

You: Alt+S what's causing this?

Reply: KeyError: 'region' in load_config(). The staging YAML has no region key. Use cfg.get("region", "us-east-1") or add the key to staging.yaml.

Regex

You: explain in plain words: ^(?=.*\d)(?=.*[a-z]).{12,}$

Reply: At least 12 characters, with at least one digit and at least one lowercase letter anywhere in the string.

One monitor

You: Alt+Shift+2 p95 latency jumped at 14:05. which service?

Reply: The checkout-api panel climbs from about 180 ms to 900 ms at 14:05. The other services stay flat, so start with its last deploy.

Shell

You: PowerShell one-liner: files over 1 GB under C:\Users

Reply: Get-ChildItem C:\Users -Recurse -File -ErrorAction SilentlyContinue | Where-Object Length -gt 1GB | Select-Object FullName, Length

Setup

Set it up for coding

Install

Run the 50 MB installer on Windows 10 or 11 and paste your license key.

Connect a model

Sign in with GitHub to use Copilot, or add a key for OpenAI, Claude, Gemini, Azure, or a localhost endpoint.

Write a dev system prompt

For example: "Senior C# and TypeScript dev. Answer with code first, explanation after, max 5 lines."

Dock it

Park it beside the editor at 80% opacity. Alt+C lets clicks pass through to the code underneath.

Compare

IDE assistant, browser tab or an AI coding assistant overlay

The two work well side by side. Here is what each one covers.

IDE assistant (Copilot in VS Code, Cursor)ChatGPT in a browser tabAuraCloak
Reads your project filesYesNoNo, send text files or screenshots
Inline completions and editsYesNoNo
Sees terminal, browser, RDP, PDFsNoManual uploadAlt+S screenshot
Stays on screen while you work elsewhereOnly inside the editorNoAlways on top, click-through
Shows on your screen shareYes, as part of the editorYes, if on the shared screenNo, excluded from capture
Local modelsSome editors allow itNoOllama, LM Studio, vLLM

Ask about any window, not just your editor

$9 a month or $49 once. Every feature in both plans.

FAQ

AI coding overlay questions

Does AuraCloak replace GitHub Copilot in my editor?

No. AuraCloak has no inline completions and doesn't edit files. It is a chat overlay for questions about anything on your screen. You can sign in with GitHub inside AuraCloak to use the models on your Copilot subscription in the overlay too.

How does the AI see my code?

Through what you send it. Press Alt+S to screenshot all monitors or Alt+Shift+1 to 9 for one monitor, paste an image with Ctrl+V, drag in text files, or paste the code as text. AuraCloak does not read your project folder on its own.

Can I keep proprietary code off third-party servers?

Yes, with a local model. Add a custom OpenAI-compatible endpoint such as Ollama at http://localhost:11434/v1/chat/completions or LM Studio, and prompts never leave your PC. With cloud providers, prompts go directly from your PC to that provider and not through AuraCloak servers.

Will my pairing partner see the overlay when I share my screen?

Not through the share. AuraCloak sets the Windows WDA_EXCLUDEFROMCAPTURE flag, so Zoom, Teams, Slack, Discord and OBS leave it out of what they capture. It needs Windows 10 version 2004 or later. You can turn this off in Settings if you want to show the overlay.

Which models can read screenshots of code?

Vision-capable models: OpenAI gpt-4o and later, Claude 3 and later, all Google Gemini models, gpt-4o through GitHub Copilot or Azure OpenAI, and custom endpoints that accept the OpenAI vision format.