avast

avast / sage

Public

Lightweight Agent Detection & Response (ADR) layer for AI agents — guards commands, files, and web requests

ai.gendigital.comsage agents ai claude-code claude-code-plugin cursor-ai
84
5
100% credibility
Found Feb 19, 2026 at 29 stars 3x -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
TypeScript
AI Summary

Sage provides a security layer for AI coding agents in tools like Claude Code, Cursor, VS Code, and OpenClaw by intercepting and vetting tool calls against threats.

How It Works

1
🔍 Discover Sage

You learn about Sage, a friendly safety shield that protects your AI coding helper from risky actions like bad websites or dangerous commands.

2
🛠️ Add to your AI tool

Pick your favorite AI coding app like Claude Code or Cursor, and add Sage with one easy step.

3
✅ Turn on protection

Click 'Enable Protection' and Sage quietly starts guarding your sessions without getting in the way.

4
🤖 Chat and create

Talk to your AI as usual, asking it to run commands, fetch info, or edit files.

5
Sage spots something
👍
All safe

Everything passes smoothly and your work continues.

⚠️
Risk flagged

Sage highlights the suspicious part and asks if you approve.

6
👌 You decide

If it's safe, you give the okay; Sage remembers for next time.

🛡️ Secure coding bliss

Your AI helper works faster and safer, threats blocked, and you focus on building great things.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 29 to 84 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is sage?

Sage delivers a lightweight Agent Detection & Response layer for AI coding agents, scanning Bash commands, file ops, web fetches, and plugins for threats like malicious URLs, supply-chain risks, and obfuscated patterns. TypeScript-based, it hooks into Claude Code plugins, Cursor/VS Code extensions, and OpenClaw to enforce verdicts (block/ask/allow) via local YAML heuristics, cloud rep checks, and package age analysis—keeping your dev environment secure without heavy EDR overhead.

Why is it gaining traction?

Unlike full-blown CrowdStrike-like agents, Sage is a connect-secure lightweight agent that installs in seconds via plugin marketplaces, runs offline-first, and flags GitHub lightweight agent risks like fresh npm/PyPI packages or piped shell tricks—perfect for the agentic AI boom where tools like lightweight ai agent frameworks execute unvetted code.

Who should use this?

Developers wielding AI agents in Cursor, Claude Code, or VS Code for code gen/installs, especially those hit by supply-chain scares or needing a lightweight coding agent guard in terminals—ideal for solo hackers or small teams prototyping lightweight agentic ai without enterprise bloat.

Verdict

At 18 stars and 1.0% credibility, Sage feels alpha but packs thorough docs, session-start scans, and MCP tools for allowlisting—grab it now if you're early on lightweight ai agent frameworks, but monitor for maturity as adoption grows.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.