BugAI
AI-powered bug detection and one-click fixes, running entirely in your editor on Groq's ultra-fast inference.
Features
- Live analysis - edited lines are analyzed automatically as you type; files are re-checked on save.
- Liquid line markers - bugs glow red behind the line, suggestions glow yellow. A colored lens sits in the gutter next to every flagged line.
- One-click fixes - hover a flagged line or open the notification panel and hit Fix. BugAI proposes the corrected code and applies it as an undoable editor edit.
- Notification panel - a glass notification strip docks in the bottom panel area, listing every issue with its line number, a Fix button and a dismiss cross. Click any notification to jump to the line.
- Problems integration - every finding is also pushed to the Problems panel and the editor overview ruler.
Getting started
Install the extension, open a file and wait a moment. The status bar shows BugAI's state: a spinner while analyzing, the issue count, or a checkmark when the file is clean.
Manual triggers:
BugAI: Scan Active File - full analysis of the current fileBugAI: Scan Workspace - batch analysis of up to 60 source filesBugAI: Clear All Issues - wipe current findingsBugAI: Open Notifications - focus the notification panel
Configuration
| Setting |
Default |
Description |
bugai.model |
llama-3.3-70b-versatile |
Groq model used for analysis |
bugai.apiKeys |
[] |
Optional Groq API keys. When empty, the built-in key pool is used |
bugai.autoScan |
true |
Analyze edited lines and re-analyze on save |
bugai.maxLinesPerScan |
800 |
Maximum lines sent per analysis request |
For team deployments, configure bugai.apiKeys with per-developer keys to keep usage on your own account.
Requirements
Visual Studio Code / Codium 1.85.0 or newer. No other dependencies.
Web support
BugAI runs in the browser too: vscode.dev, github.dev, OpenVSCode Server and any other VS Code for the Web host. The extension ships a dedicated browser build (browser entry) with no Node.js dependencies, and Groq's API is CORS-enabled so analysis works directly from the browser.
To try it locally in a web environment:
# VS Code desktop, serves the web build at http://localhost:8080
code serve-web --extensionDevelopmentPath=/path/to/bugai
Note: vscode.dev and github.dev can only install extensions published to a marketplace, so a published BugAI (or an OpenVSCode/VSCodium server that loads the vsix from disk) is needed there.
Release notes
See CHANGELOG.md.
License
MIT
