Glossary · Trina AX Docs

Term	Meaning
Activity log	Admin records of actions and tool calls. Shows which files the AI read, never the chat.
AI connection	The account an engine uses: a Claude subscription, an API key or a ChatGPT sign-in.
App connector	The Team AX MCP connector: your own Claude connecting to a team.
Confidential chat	A chat that never uses the web or sends sound out (with Safe operations 1A on).
Core	Team AX, the shared codebase Trina AX is built on.
Engine	The agent that answers: Claude Code, Codex or the API engine.
Group	A set of people that folder rules can name.
Local-login	Engines use the server OS user's own CLI sign-in. Used by office Macs and development.
Managed	Engines use AI connections through the model proxy. Used by Linux servers.
MCP	Model Context Protocol. How the AI talks to tools and outside services.
Model proxy	The server part that adds real keys to model calls, and counts usage.
Owner	The one person with every admin power plus team-level choices.
Plugin	A package admins install: a skill bundle or an isolated native tool.
Project	A team folder with instructions. Chats in it get that folder and those instructions.
Relay	The service that gives a team server a public address through outbound connections.
Scheduled task	A request the AI repeats on a schedule, as the person who made it.
Skill	A `SKILL.md` file: a reusable procedure. Never extra access.
Skill manager	A person who can change team skills, through write access to `skills/`.
Sub-agent	A helper the AI starts for part of a big request, working as the same person.
Team folder	The folder tree on the server: `shared/`, `groups/`, `projects/`, `skills/`, `users/`.
Turn	One answer from the AI, with all its tool calls.