Open Source · MIT License

Stop your AI agent before it breaks production

Pre-execution security guards for AI coding agents. Every command is checked before it runs. Dangerous operations are blocked, not logged after the fact.

$ npx guardrail-agent init View on GitHub

Works with Claude Code · Cursor · GitHub Copilot · Windsurf

agent $ rm -rf /etc/nginx BLOCKED destructive path
agent $ git push origin main BLOCKED protected branch
agent $ env BLOCKED exposes all environment variables
agent $ DELETE FROM users BLOCKED no WHERE clause
agent $ curl webhook.site/token BLOCKED exfiltration domain
agent $ iptables -F BLOCKED firewall flush
agent $ git push origin develop ALLOWED
agent $ npm install express ALLOWED
Open Source

10 core guards, zero configuration

Install once. Every command your AI agent runs is checked against production-grade security rules.

pre-bash

main_push_guard

Blocks direct push to main/master, force push, and git reset --hard.

pre-bash

basic_pii_gate

Blocks env, printenv, /proc/environ, docker inspect without --format.

pre-bash

basic_secret_detector

Blocks exfiltration to webhook.site, requestbin, pipedream, and similar.

pre-bash

destructive_path_guard

Blocks rm -rf on /etc, /usr, /var, /home, and other protected paths.

pre-bash

firewall_flush_guard

Blocks iptables flush, ufw disable, nft flush ruleset.

pre-bash

service_protection_guard

Blocks stop/disable/kill on sshd, docker, postgres, nginx.

pre-bash · database

mass_update_guard

Blocks UPDATE/DELETE without WHERE on protected tables.

post-bash

env_dump_detector

Detects full environment dumps (8+ KEY=VALUE lines) in output.

post-bash

basic_injection_scanner

Detects prompt injection like "ignore previous instructions" in output.

post-edit

error_swallow_guard

Catches empty catch blocks in payment, webhook, and cron code.

100+
Guards in production
15
Apps protected
41
Regression tests
0
False negatives
GuardRail Pro

40+ guards from real incidents

Every Pro guard exists because something went wrong in production. These are patterns you don't design on a whiteboard.

Advanced PII Detection

15+ leak vectors beyond env/printenv. Script content analysis. SSH wrapping detection.

Multi-Step Attack Defense

Detects coordinated attack sequences across commands. Catches what single-command guards miss.

Supply Chain Security

npm audit integration. License compliance. Blocks packages with known vulnerabilities.

Self-Bypass Prevention

Detects AI agents attempting to disable or circumvent their own security guards.

PEN-Test Framework

50+ attack patterns. Profile escape testing. Bypass persistence checks. JSON report.

Compliance Reports

EU AI Act article mapping. Risk classification. PDF audit reports for regulators.

€20 /dev/month
Full guard suite, PEN-test framework, priority support
Contact for Pro
€5.000 one-time
Compliance kit: audit reports, article mapping, PDF export
Contact for compliance
EU AI Act enforcement begins August 2, 2026

Compliance is not optional

If your team uses AI coding agents in the EU, you need auditable governance. GuardRail provides pre-execution controls with a full audit trail.

Article 9

Risk Management

Guard classification. PEN-test framework. Risk-based guard activation per command profile.

Article 10

Data Governance

PII gates prevent secret exposure. Environment dump detection. Container inspection controls.

Article 12

Record-Keeping

Timestamped audit log. Every blocked command recorded with guard name, reason, and context.

Article 14

Human Oversight

deny() gates block before execution. Approval workflows for critical operations.

Article 15

Accuracy & Robustness

Injection defense. 41 regression tests. True positive and false positive checks per guard.

Get started in 30 seconds

One command. 10 security guards. Zero configuration.

$ npx guardrail-agent init GitHub
Documentation Report a bug Contact