peg

peg / rampart

Public

Open-source firewall for AI agents. Policy engine that controls what OpenClaw, Claude Code, Cursor, Codex, and any AI tool can do on your machine.

rampart.sh agent-security ai-agents ai-security audit-trail claude-code
30
5
100% credibility
Found Feb 17, 2026 at 21 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Go
AI Summary

Rampart is an open-source security layer that monitors and controls dangerous actions by AI coding agents like Claude Code and OpenClaw, blocking risky commands and providing tamper-proof audit logs.

How It Works

1
🔍 Discover Rampart

While using AI coding helpers like Claude or OpenClaw, you learn about Rampart to keep them from running risky commands on your computer.

2
⚡ Install in seconds

Run a simple install command and Rampart is ready—no complicated setup needed.

3
🔗 Connect your AI helper

Tell Rampart which AI tool you're using, like Claude Code, and it starts watching every action.

4
👀 Watch it work live

Open the dashboard to see safe actions zoom by in green and dangers get blocked in red.

5
📊 Check your safety record

Review the secure log of everything your AI tried to do, with proof nothing sneaky slipped through.

🛡️ AI safely supercharged

Your AI helper now builds and codes confidently while Rampart keeps your machine protected.

Sign up to see the full architecture

4 more

Sign Up Free

Star Growth

See how this repo grew from 21 to 30 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 rampart?

Rampart is an open-source firewall software for AI agents, written in Go, that lets you define YAML policies to control what tools like Claude Code, Cursor, OpenClaw, or any MCP server can execute on your machine. It intercepts shell commands, file reads/writes, and network calls via native hooks, shell shims, LD_PRELOAD, or proxies, blocking destructive actions like `rm -rf /` while logging suspicious ones and requiring approval for deploys. Install via Homebrew or Go, then `rampart setup claude-code` for instant protection—even in dangerously permissive modes.

Why is it gaining traction?

Unlike generic open source firewalls for Linux, Windows, or routers, Rampart targets AI coding agents specifically, with one-command setups for popular tools and microsecond policy evaluation that doesn't slow down your workflow. Features like live TUI dashboards (`rampart watch`), hash-chained audit trails, and webhook notifications stand out for devs needing visibility without overhead. As a self-hosted github copilot alternative, it fills a gap for safe AI experimentation.

Who should use this?

AI-assisted developers running Claude Code or Cursor who worry about agents nuking repos or exfiling keys. DevOps teams deploying unsupervised infra agents via OpenClaw or MCP servers, needing policy enforcement and SIEM integration. Anyone evaluating open source github copilot alternatives for local, controlled coding.

Verdict

Try it if you're using AI agents daily—solid docs, CLI like `rampart approve`, and community policies make onboarding fast despite 19 stars and 1.0% credibility score. Still early (low test coverage on edge cases), but promising for paranoid setups; watch for stability as it matures.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.