Using workspace knowledge.

Workspace knowledge is a shared pool of named text notes that every workspace chat's AI reads before it answers. Think of it as a small, always-attached briefing document — no member has to paste it in themselves, no member has to remember it exists.

Common uses:

- Brand voice guide. - Glossary of internal terms (e.g., how you name products, teams, projects). - The company's positioning: what you sell, who you sell to, how you talk about competitors. - A house style guide (formatting rules, forbidden words, must-say phrases). - Escalation SOPs: when to route to which team, what tone to use with which customer segment. - Product-line facts the assistant might otherwise get wrong (prices, features, availability).

To manage: open Workspace settings → **Knowledge**. Only admins can add, edit, or delete notes. Every workspace member can read the list.

Each note is up to 20,000 characters. The AI concatenates newest-first up to a ~24,000-character budget across the whole knowledge pool, so keep notes focused. If you go past the budget, older notes get dropped from that turn's context.

Notes are added directly to the AI's system prompt, not to your chats. That means:

- The assistant treats them as facts. Don't put unverified information in there — the assistant won't caveat it. - The assistant is asked to cite the note's title inline when it uses a note directly ("per the Glossary note, APAC = Asia-Pacific"). This is a soft rule, not a guarantee. - The notes are visible to every active member of the workspace, at any time, via the Knowledge page.

**Not for**: multi-page attachments (upload those to a specific project instead), member-specific configuration (put that in each member's Behavior settings), sensitive customer data (workspace knowledge is workspace-wide — everyone sees it).

If you remove a member from the workspace, they lose access to the knowledge pool on their next request. If you delete a note, the AI stops using it on the next turn.