Built-in tools›Long-term memory

Group

Long-term memory

Turned on by

The memory switches on the Advanced tools page of Conversation settings.

Approval

Not reviewed: runs without asking, within the limits on this page.

Profile key

tool.read_project_memory.description

Reads one document from the project memory, the notes the model keeps for itself about one of the conversation's workspaces.

Each workspace has its own project memory, in .mework/memory/ inside the workspace's folder. While Enable project memory is on (Conversation settings → Advanced tools), each workspace's index MEMORY.md is in the model's context on every run, and the model reads a document with this tool when an index suggests it is relevant. See Memory and instructions.

Limits

• A document over 256 KiB cannot be read.

Good to know

• In a conversation with more than one workspace, the tool takes an optional workspace number saying whose memory to read (default 1). With one workspace it has no such parameter.
• A document created during a run appears in the index from the next run on.

Parameters