Get started What Oxelone does
What Oxelone does
Oxelone is an AI security posture management platform. It helps a security team answer one question in minutes: what AI do we have, and how exposed are we? It also produces the evidence an auditor asks for.
It is built for security and compliance teams first. AI and machine learning engineers who own the models, pipelines and agents use it too, and so do executives who want posture summaries rather than individual alerts.
What it keeps track of
Oxelone builds an inventory of the AI in your workspace, from two kinds of collection.
- The agent. A small program that runs on a machine. It reports the AI tools, models, skills and MCP configurations it finds there, and it can capture the interactions people and tools have with AI services. See The agent.
- Connectors. A connector reads a platform's own API for AI posture: its users, settings and records. GitHub, Microsoft 365, Google Workspace, OpenAI and Anthropic each have one. Nothing is installed on the platform. See Connectors.
Both feed the same inventory, so a tool found on a laptop and a user found on a platform show up in the same place.
What you can do with it
| You want to | Where it lives |
|---|---|
| See every machine that runs the agent | Inventory › Machines |
| See the AI artifacts found on those machines: tools, skills, MCP configurations, models and packages | Inventory › AI Artifacts and Inventory › AI Tools |
| See the people behind the AI use, merged across machines and platforms | Inventory › Accounts |
| See the platforms you have connected | Inventory › Connections |
| Read the interactions with AI services that your workspace keeps | Activity › Interactions |
| Write a rule about what is allowed | Governance › Policies |
| See what each rule decided | Governance › Runs |
| Triage what needs a person | Detections › Alerts and Detections › Issues |
| Read your posture by AI risk | Governance › Posture |
Rules, engines and what they produce
A policy says what to watch for and what to do about it. It has a trigger (a network interaction, a file on a machine, an action an AI tool is about to take, or a record on a connected platform), a filter and a list of actions. See Policies.
A policy can ask an engine a question about content: does this text hold a secret, is this an attempt at prompt injection. The engine answers and the policy decides. See Engines.
When a policy matches, it can:
- change what happens, for example block, ask, redact or allow what a person or an AI tool sends;
- raise an alert, a record that something happened;
- open an issue, a record of something that stays true until it is fixed.
Alerts and issues are different things and are triaged differently. See Alerts and issues.
Try a rule before it acts
A new policy starts disabled, on no machine. Its Monitor state records every match and raises its alerts and issues, but it stops nothing and changes nothing. Use it to see what a rule would catch on real traffic before you let it act. Rolling out shows the order to do it in.
Where to go next
- Your first hour takes an empty workspace to its first alert.
- Concepts and glossary defines the words the console uses.
- A tour of the console shows where everything is.