Open vs Enterprise

Documentation Index

Fetch the complete documentation index at: https://docs.agentguardian.io/llms.txt

Use this file to discover all available pages before exploring further.

AgentGuardian comes in two layers. They share the same attack engine and the same canonical scan.json, but they solve different problems. These docs cover only the open-source layer.

​

AgentGuardian (open source)

AgentGuardian is an open-source red-teaming toolkit for AI agents. It is licensed Apache-2.0, runs fully local, and is designed to drop into a developer’s loop and into CI/CD. It is not a runtime gateway. It is not a guardrail product. It is not a policy proxy. It is the developer’s pre-flight check: write an agent, scan it, read the report, fix the finding, commit the fix. Use AgentGuardian when you want to:

• Find prompt injection, tool abuse, RAG poisoning, memory poisoning, cascading failures, and unsafe agent behaviour before you ship.
• Run reproducible adversarial scans in CI on every PR.
• Generate SARIF for GitHub Code Scanning, JUnit for any CI dashboard, PDF for an internal review, Markdown for a PR comment.
• Cite a deterministic AIVSS score in a security review.
• Embed adversarial testing in your own pipeline via the Python SDK.

You get all of that for free, forever, under Apache-2.0. No telemetry is collected unless you explicitly opt in with agent-guardian telemetry on. Every scan stays on your machine; no findings, prompts, or transcripts leave your laptop.

​

AgentGuardian Enterprise

AgentGuardian Enterprise is the commercial layer Glacien ships on top of the same open-source engine. It is not documented in these docs — by design. It exists to solve the problems that show up after a security team adopts AgentGuardian across many agents and many teams:

• Managed evidence packs — a central, signed, queryable store of every scan run by every team, with retention, search, and audit trails.
• Team workflows — assignment, triage, finding lifecycle, SLA tracking, escalation, integrations with Jira / Linear / ServiceNow.
• Audit dashboards — historical AIVSS trends, attack-surface coverage maps, evidence-pack browsing for auditors and risk teams.
• Policy governance — declarative agent-safety policies that fail a release if a high-risk finding regressed, with enforcement at the PR-check layer and at the runtime layer.
• Runtime controls — the governance and enforcement plane that sits in front of production agents and the only place AgentGuardian acts as a defensive runtime. Open is not that.
• Commercial support from Glacien — SLAs, dedicated engineering contacts, custom adapter and probe development.

Both layers consume the same canonical scan.json and the same probe corpus. Findings produced by AgentGuardian can be ingested by Enterprise without re-scanning. Adapters and strategies developed upstream in the open-source project flow into both.

​

Quick reference

​

A note on positioning

These docs are the developer-first home of the open-source project. They will not turn into a sales funnel. The Enterprise page is a single cross-link to glacien.com for readers who arrive here and need the commercial layer. Everything else in this site is the OSS toolkit — what it does, how it works, how to run it.

​

Where to go next

• Quickstart — install and run your first scan in five minutes.
• How AgentGuardian works — the four-step developer mental model.
• Reports overview — what a scan produces.
• Enterprise — the one cross-link page for commercial features and Glacien support.