Sentinel is an AI-powered extension that automatically analyzes your code after every edit.
Features
Easy setup and customization: Choose which files to track using a .sentinelignore file (similar to .gitignore), define custom detection logic in an instructions.md file, and select your preferred LLM model.
Smart issue detection: View issues directly in a sidebar, navigate to exact problem lines, and resolve them efficiently.
Context-aware analysis: Sentinel analyzes both local file context and global codebase context for more accurate and relevant issue reports.
Automatic updates: Each time an edit is made, Sentinel will automatically run in the background, analyzing your code for bugs and other criteria, as specificed in instructions.md.
CLI commands: Run gather, analyze, and init from your terminal.
Usage dashboard: A hosted webpage shows your per-model token usage, trends, and how close you are to your usage limits (open quickly via sentinelai status).
Setup
Choose and configure supported LLM models. In VS Code use: "Sentinel: Select Gather Model" and/or "Sentinel: Select Analyze Model". For CLI, list models with sentinelai models, then select with sentinelai select-model <id> or set individually with sentinelai select-gather-model <id> / sentinelai select-analyze-model <id>.
Create starter files — this generates .sentinelignore and instructions.md for you. Use the Command Palette: "Sentinel: Create Starter Files" or run sentinelai init-files.
Sign in — no API keys. Use OAuth.
VS Code: run "Sentinel: Sign In" (opens the hosted web app). After login, the site redirects to the provided VS Code callback and the extension stores your token.
CLI: run sentinelai login. This opens the same web app in CLI mode with a hidden token field. After signing in, click Reveal and Copy, then paste into the terminal prompt. The token is validated and saved for future sessions.
Initialize Sentinel — this runs a gather step to determine the proper global context for each active file, then analyzes them. Use the Command Palette: "Sentinel: Begin Initialize" or run sentinelai init.
Edit as you normally would — Sentinel will automatically handle the rest, analyzing each time an edit is made.
If a file or its related context changes significantly, re-gather the global context for that file. Use the Command Palette: "Sentinel: Gather Global Context" or run sentinelai gather <file>. To re-run issue analysis without re-gathering, use sentinelai reanalyze <file>.
CLI
Install: npm install -g sentinelai
Initialize all active files: sentinelai init
Create starter files: sentinelai init-files
Re-gather context for a file: sentinelai gather <file>
Analyze a file with last gathered context: sentinelai analyze <file> (or sentinelai reanalyze <file>)
Sign in: sentinelai login (opens browser and prompts to paste token), or sentinelai login <token> to paste directly. Sign out: sentinelai logout; show account: sentinelai whoami. Open the dashboard: sentinelai status.
