Getting Started
Control AI-generated code before release with policy enforcement and human approval.
What is Guardian?
Guardian helps small engineering teams control AI-generated code before it ships. It is a desktop + CLI governance layer focused on risky AI-assisted changes and release decisions.
Guardian is not:
- A generic AI coding assistant
- A generic code review tool
- A generic security scanner
Guardian is the release control layer for AI-heavy code change workflows.
Core Product Workflow
- AI-Generated Code Intake: Detect and isolate AI-assisted or unusually large/refactor changes for stricter handling.
- Policy Enforcement: Apply team architecture, security, and quality policies automatically.
- Human Approval Workflow: Require accountable approval, override owner, and reason when risk is high.
- Release Decision Surface: Answer the final release question with evidence: can this code ship now, and why?
Who is it for?
Guardian is designed for teams that need release confidence on AI-assisted code:
| Role | Use Case |
|---|---|
| Engineering Leaders | Standardize release decisions across AI-heavy PRs |
| Platform/DevOps Teams | Enforce guardrails in local + CI release gates |
| Security/Architecture Teams | Apply non-negotiable policies before merge/release |
| Release Managers | Track approvals, overrides, and decision evidence |
How it works
Guardian follows a four-step release governance cycle:
┌─────────────┐ ┌──────────────┐ ┌──────────────┐ ┌─────────────┐
│ INTAKE │────▶│ ENFORCE │────▶│ APPROVE │────▶│ DECIDE │
│ AI-heavy PR │ │ Team Policy │ │ Human Gate │ │ Ship/Block │
└─────────────┘ └──────────────┘ └──────────────┘ └─────────────┘
- Intake: Guardian identifies risky AI-assisted or unusually large changes.
- Enforce: Team policies are evaluated across architecture, security, and quality rules.
- Approve: Human reviewers validate fixes and document approval or override reasons.
- Decide: A release readiness signal is generated with explainable evidence.
First 10 Minutes
Here's what you should do in your first session:
1. Install the Application
Download the installer from the Download page. macOS builds are available now; Windows and Linux appear when published.
2. Configure Provider Settings
Navigate to Settings > Provider and configure your AI provider credentials. Guardian supports both cloud and local models.
3. Start Monitoring Your Workspace
Click the Scope field (or drag and drop a folder onto the app) to select your workspace, then choose Launch Guardian to start monitoring.
4. Review Findings as a Release Decision
Once the scan completes:
- Focus first on AI-heavy and architecture-sensitive changes
- Use Guru to understand why each risky change matters
- Apply policy-aligned fixes and record decision context for your release flow
Application Layout
The Guardian interface consists of four main areas:
| Area | Purpose |
|---|---|
| Sidebar | Navigate monitoring, reviews, settings, and history |
| Main Panel | Inspect findings, policy evidence, and fix proposals |
| Guru Chat | Explain risky changes and propose policy-aligned fixes |
| Status Bar | Track provider status, version, and workflow state |
Next Steps
- Installation Guide - Detailed setup instructions
- Guru - Explain risky findings and generate fix proposals
- Reviews & Fix Proposals - Manage approvals and controlled apply flow
- Configuration - Configure policies, scan scope, and guardian.lock
- Security - Understand the local-first security model