Get started Concepts and glossary

Concepts and glossary

Oxelone uses a small, fixed set of words. The same word means the same thing on every screen and on every page of these docs. This page lists them.

Your workspace

Term	What it means
Workspace	An isolated customer boundary. Your machines, connections, policies and records all belong to one workspace.
Slug	A workspace's short name. The console keeps the active workspace in your session, not in the address, and the workspace switcher changes it.
Plan	What a workspace is allowed: a limit for each meter and how far interaction logging may go. It is shown in Settings › Plan & licence.
Meter	One thing a plan limits: Machines, Members, Connections, Interactions, Engine runs or Billable units.

What collects the data

Term	What it means
The agent	The program that runs on a machine and reports what it finds. Its command is `oxel-agent`.
Companion	The tray app beside the agent. It is what a person at the machine sees.
Machine	A computer that runs the agent. It has a record under Inventory › Machines.
Enrollment profile	A saved set of settings for machines, including which capture strategies they run. A machine takes its profile when it enrolls.
Capture strategy	One way the agent collects: TLS interception, Network steering, CLI & IDE tools, Browser extension or Filesystem scan. See Capture strategies.
CLI & IDE tools	Command-line and editor tools that use AI, such as coding assistants. The agent governs them through the tool's own hook.
Hook	A command a CLI or IDE tool runs before it acts, which the agent answers.
Removal key	A key issued for one machine that authorizes removing its agent on that machine.
Connector	The way Oxelone reads a platform's API for AI posture, such as GitHub or Microsoft 365.
Connection	One configured connector in your workspace, with its own credentials, sync period and rules. You can have many for one connector.
Records	What a connector reads: record kinds such as Users, Repositories, Files, Audit events and Settings.
Sync	One run of a connection's collection. Sync now starts one.
Readable / Not readable	Whether a connection's credential reaches a kind of record, always with the reason.

What it finds

Term	What it means
AI artifact	Something that belongs to AI use and is recorded in the inventory: a tool, a skill, an MCP configuration, a model, a package.
Category	The main kind of an AI artifact, such as AI Tool, Skill, MCP, Model Weights or AI Dependency.
AI tool	An AI artifact in the AI Tool category, for example a coding assistant. Inventory › AI Tools lists them.
Location	Where an artifact is.
Person	One row on Inventory › Accounts: a merged identity, built from an address, a platform login or an operating-system account.
Interaction	One intercepted exchange between a person or tool and an AI service. Activity › Interactions lists them.
Content	The text of an interaction: the prompt, the response and any attachment. Your workspace's logging mode decides whether it is kept.
Attachment	A file a person attached to an interaction.
Session	One run of an AI tool, read from its transcript. Only metadata is kept.

What you decide

Term	What it means
Policy	A trigger, a filter and a list of actions. It says what to watch for and what to do about it.
Network, Host, Hook, Connector	The four kinds of policy trigger: an interaction, a file on a machine, an action an AI tool is about to take, and a record on a connected platform.
Response	The one thing a policy does that changes something, such as Allow, Ask, Redact, Block or Lock. A policy has at most one.
Report	What a policy writes into the console: Raise an alert, Open an issue or Tag the artifact. These run in Monitor too.
Disabled, Monitor, Enabled	A policy's three states. Disabled puts it on no machine. Monitor records every match and raises alerts and issues but stops and changes nothing. Enabled takes its actions.
Group	A named, ordered set of policies of one type with one On / Off switch.
Provided by Oxelone	The mark on anything Oxelone supplies that your workspace did not write.
Engine	A built-in checker a policy can ask about content, such as PII and secrets or prompt injection. An engine answers; it never decides.
AI judge	The engine that grades content against a rubric you write, using an AI model.
Default model	The model that every judge and built-in scanner asks unless it names its own.

What it tells you

Term	What it means
Alert	A record that something happened, raised by a policy. It is shown under Detections › Alerts.
Issue	Something that is still true, such as a setting that leaves AI exposed. It is opened, acknowledged and resolved under Detections › Issues.
Decision	One row of Governance › Runs: what a policy decided about something, blocks included.
Notice	A short note the agent shows a person at their machine. It is not stored. It is not an alert.
Posture	The Governance page that reads every shipped posture check by AI risk.
Posture check	A rule that reads settings or metadata and decides.
Inbox	The list of notifications behind the bell in the top bar.
Inspector	The record docked beside a list when you open a row.

A machine's life

Term	What it means
Installing	Enrolled, and no heartbeat yet.
Active	The agent has checked in.
Stuck	Enrolled, and never checked in after 15 minutes.
Suspended	A reversible stop: the agent puts back what it changed and goes quiet. Resume restores it.
Uninstalled	The agent was removed on the machine with its removal key. The record is kept for its history.
Archived	A Stuck or Uninstalled machine, hidden from the default list after 30 days.
Delete	The final step. It is offered for a machine that is Suspended, Stuck, Installing, Uninstalled or Archived.

Getting help

Term	What it means
Ticket	What a person opens to tell Oxelone something: a Bug, an Idea or a Question.
What's new	The console's changelog, in the Help menu.
Announcement	A notice from Oxelone's platform admins to workspaces: News, Maintenance, Incident or Action required.