Memory
Durable, plain-text context your team and your AI agents both read and write — conventions, decisions, and definitions that should outlive any chat window.
On this page
What memory is for
Memory is shared context that people and agents both use. Keep the things an agent should know before it acts: conventions for naming and prioritizing work, your definition of done, who owns what, and decisions worth remembering across sessions. It is plain text, searchable, and attributed — and it is available through the same MCP tools agents already use.
Creating a memory
On Memory, click New memory and give it a title, a short description, and plain-text content. The title and description are how people and agents find it in a list; the content is the full note.
Editing and deleting
Open a memory to edit the title, description, and content in place — changes auto-save as you type. Deleting a memory removes it for everyone and every agent, so use it when a note is genuinely no longer true.
Who can read and write
Everyone in the workspace, including viewers, can read memory. Creating and editing needs the member role or higher — the same bar as changing other workspace content.
How agents use memory
Over MCP, agents list and read memories at the start of a task and write to them when they learn something worth keeping. The relevant tools are list_memories, get_memory, create_memory, update_memory, and delete_memory.
Tip
Write memory the way you'd brief a new teammate: short, specific, and about how your team works — not secrets or one-off data.
Improve this article
Spot something missing, outdated, or a broken link? Send the docs team a note.