Public Beta • Free 1,500 cloud operations/mo* (Local checks are offline & unlimited)

Stop AI agents from breaking production

Preview diffs and risk scores in Slack. Block dangerous operations before they run.

⚠️ Disclaimer: This is an experimental tool. It handles sensitive API tokens using AES-256-GCM encryption. Use at your own risk.

How to Setup

1

Install CLI

npm install -g @saferun/cli
2

Run Setup

Navigate to your repo and run:

saferun setup
3

Get API Key

Enter your email to get API key:

Free 1,500 cloud ops/mo*

Here is how it should work

Who needs this

✓
Developers using AI agents (Cursor, Claude Code, Windsurf) — Those who click "Always allow" and want to sleep peacefully.
✓
"Vibe-coders" and Non-Senior Devs — Those who trust AI logic but want a "second opinion" before a fatal action.
✓
Teams working on critical repos — Where one accidental push --force can paralyze everyone for hours.
✓
AI Agent Developers — Looking for a ready-to-use Git safety protocol for their autonomous bots.

How it works

Three layers of protection

1️⃣

Shell Wrapper

Intercepts destructive commands locally before execution. (Offline & Zero-latency)

2️⃣

Core Git Hook

Catches direct /usr/bin/git calls via Git 2.29+ hooks

3️⃣

GitHub Webhooks

Real-time Slack alerts & one-click recovery for remote accidents (requires GitHub App)

Protected Operations

⛔
History Rewrites — git rebase, git reset --hard (detected via state-change)
⛔
Force Pushes — git push --force, --force-with-lease
⛔
Branch Deletion — git branch -D
⛔
Protected Branches — Risk-based alerts for direct commits to main or master
For early adopters: You're getting free access during beta. Your feedback shapes the product. When we add payments, you'll get lifetime discount.
→ Request features or report issues on GitHub

Links

API DocsGitHubnpmAuthor