Memory Tools

Source: src/langgraph_kit/core/tools/memory_tools.py

Agent-callable tools for managing persistent memory through the PersistentMemoryManager.

build_memory_tools(memory_manager)

Returns a list of 5 async tool functions:

save_memory(title, type, scope, summary, body)

Create a new persistent memory record.

Parameter	Type	Description
title	str	Short descriptive title
type	str	Memory type: user, feedback, project, reference
scope	str	Memory scope: user, assistant, project, team
summary	str	One-line summary
body	str	Full content

list_memories(scope, type=None)

List stored memory records by scope, optionally filtered by type.

search_memories(query, scope)

Semantic search for relevant memories within a scope.

update_memory(id, scope, body=None, summary=None)

Update an existing memory record. Only provided fields are changed.

delete_memory(id, scope)

Delete a memory record by ID and scope.

Registration

Memory tools are typically registered via the builder utility:

from langgraph_kit.core.graph_builder.tools import register_memory_tools

register_memory_tools(tool_registry, memory_manager)

This wraps each function as a ToolCapability with appropriate metadata:

• save_memory and update_memory → ToolRisk.MUTATING
• delete_memory → ToolRisk.DESTRUCTIVE
• list_memories and search_memories → ToolRisk.READ_ONLY