Mentions
Pull a project, a transcript, a model or a skill into your prompt with one keystroke.
Two keys do most of the work in the composer. @ brings something into the prompt;
/ runs a skill. Both open a menu you can filter by typing, and both work the same way in
every composer that has them.
What @ brings in
Projects
Mentioning a project attaches it — its knowledge and its skills load into the conversation. This is the one with real consequences; see below.
Transcripts
Hands Endless a specific transcript to read, instead of making it go looking for the right one.
Models
Names the image or video model you want used for that request.
Tools
"Generate image", "Generate video" and "Social radar" —
"Real-time briefing" — as @image, @video and @radar. Each one only
appears if your workspace has that feature on.
Social handles
Type an @ and a handle and Endless offers it per platform — Instagram, X,
TikTok, YouTube, LinkedIn, Facebook. You pick which one you meant, because
the same handle exists everywhere.
People aren't mentionable. The @ menu is for things — projects,
transcripts, models, tools and social accounts — not for teammates. There's no
way to notify a colleague from the composer.
Mentioning a model doesn't switch models
A model mention is an instruction for that request: use this one to make the image or the video. The model that powers the conversation is the separate "Select model" control, and it stays whatever you set it to.
What / runs
Typing / opens "Insert a skill" — a "Skills" list of the skills available to you,
plus "Manage skills" and "Add skill" at the end. Only skills that are turned on for you
appear: an admin sets which ones the workspace has, and you choose from what's left.
What a skill is
A packaged way of working that Endless loads only when the task needs it.
The add menu
"Open add menu" shows the same catalog as a browsable list — projects, transcripts, models, tools and skills — with "Photos & files" on top for attachments. Its search covers all of them at once.
Attaching a project, and detaching one
Mention it to attach it
Type @ and the project's name. A chip appears above the composer, and opening it confirms
what happened: "This project's knowledge and skills are loaded into this chat."
It stays attached
You mention a project once. From then on it's part of that conversation — every later message has its knowledge and skills loaded, and Endless can attach one itself mid-conversation when the work is clearly about it.
Remove the chip to detach it
"Remove from chat" stops the project loading, and that decision wins over the mentions already in the conversation — the project doesn't come back just because you typed its name five messages ago. Mention it again deliberately and it re-attaches.
Detaching is remembered on that device only. The attachment itself is stored with the conversation, but the fact that you removed it is kept in your browser. Open the same chat on another computer, in another browser, or after clearing site data, and the project is attached again — its knowledge and skills load as if you'd never removed it. If it matters, detach it again there too.
Why something isn't in the menu
Your workspace turned a model off
A model an admin has disabled isn't listed at all — not in @, not in
the model picker, and the chat can't reach it. It's gone, not greyed out.
Your plan doesn't cover it
A model your plan doesn't include is a different case: it is listed, dimmed, unselectable, and labelled with the plan it needs. You can see what you'd get, you just can't pick it.
You can't open it
Projects and transcripts are filtered to the ones you're allowed to read. Someone else's private work never appears in your menu.
The feature is off
@image, @video and @radar disappear when the workspace doesn't have
those surfaces enabled.
Turning models on and off, and turning surfaces on and off, both live under Settings → Governance — which needs permission to read governance, an admin by default. Most members never see those pages. See Roles and permissions.