DocsWork with agents
Memories
Memories are short notes that agents receive at the start of every turn, so you don't have to repeat a preference in each conversation. Manage them in Settings → Memories. Each environment keeps its own memories on its own machine.
Add a memory
Choose Add memory, write it in a sentence or two, and pick where it applies: all projects, or one project. You can also tell an agent "remember that…" in a thread; it saves the memory itself for that thread's project, and Settings shows it as Saved by an agent with a link to the thread it came from. To make it apply everywhere, edit it there.
What you write is treated as your own instruction. A memory an agent saved is passed on as background information instead, so an agent misled by something it read cannot turn that into an instruction for later turns. Editing an agent's memory makes it yours.
Turn memories off
Use memories turns them off for the whole environment: agents receive none and cannot save new ones. To leave them out of one thread, open the thread's menu and choose Memories → Don't use memories; choose Use memories to bring them back.
Which agents use memories
Codex and Claude Code turns receive memories. Other providers do not receive them yet. A change applies from the next turn.
If you turned on Codex's own memories in its configuration, Codex may keep what you ask it to remember in its own memory instead, where other providers don't see it. Say "remember this in Moonfold" to save it here.