Agent Desktop/Review and control/Use Memory across your workspace
Use Memory across your workspace
Search, filter and retire the notes Agent Desktop keeps across chats, while understanding what memory search needs.
Memory is a workspace-wide library of notes that can help future conversations keep context. Open it from My Account.
Memory library
Browse memories by conversation. The Agent creates and updates them; you can retire anything that should no longer be used.
Renato prefers concise weekly summaries.
Use local models for private notes.
Project launch uses the GitHub workspace.
Find a memory
Use the search field to find a note, then filter by kind. Notes are grouped by their origin chat and show their kind and date so you can understand where the context came from.
The view separates an empty library from an empty filter result. If a filter hides everything, choose Clear filters instead of assuming the notes were deleted.
Retire a note
Select a note's retire action and confirm. Retiring removes it from the active memory list. It does not delete the Space or the conversation that originally produced it.
Local models and memory search
The model that writes your answer can be local, but semantic memory search may use the Vinkius embedding service. If your workflow requires all content to stay on the computer, review whether memory embeddings are enabled before relying on a local generation model alone.
When to use Memory
Memory is useful for preferences, recurring instructions and durable project context. It is not a replacement for a source document. Ask the agent to confirm important facts from the connected service when the answer must be current.