Problem or Motivation
Currently, the memory system (read_memory / write_memory tools) works behind the scenes — only the model can read and write memories. Users have no direct way to view, search, or manage what the assistant has remembered about them, their preferences, or their project context.
Tools like Claude Code provide explicit memory commands (e.g., asking the assistant to "remember X" or viewing saved memories). A user-facing /memory command would give users transparency and control over persistent memories.
Proposed Solution
Add a /memory slash command for viewing and managing persistent memories.
List all memories:
> /memory
Saved memories:
- user-role — Senior engineer working on IaC templates
- feedback-testing — Prefer integration tests over mocks
- project-deadline — Release freeze on 2026-06-15
View a specific memory:
> /memory user-role
[user] Senior engineer working on IaC templates
User is a senior cloud infrastructure engineer, primarily works with
Alibaba Cloud ROS and Terraform...
Delete a memory:
> /memory delete user-role
Memory 'user-role' deleted.
Tell the assistant to remember something (natural language):
> I prefer using China regions for all test deployments, remember that
Memory saved: preference-china-regions
Alternatives Considered
- Rely solely on model-driven memory: The model already has
read_memory / write_memory tools, but users cannot see or manage what's been saved without asking the model to list them — this lacks transparency.
- File-based management: Users could manually browse
~/.iac-code/memory/, but this is not user-friendly.
Related Area
CLI / Interactive Mode
Additional Context
The backend infrastructure (MemoryManager, ReadMemoryTool, WriteMemoryTool, system prompt injection) already exists. This feature request is about adding a user-facing /memory command on top of it.
Problem or Motivation
Currently, the memory system (
read_memory/write_memorytools) works behind the scenes — only the model can read and write memories. Users have no direct way to view, search, or manage what the assistant has remembered about them, their preferences, or their project context.Tools like Claude Code provide explicit memory commands (e.g., asking the assistant to "remember X" or viewing saved memories). A user-facing
/memorycommand would give users transparency and control over persistent memories.Proposed Solution
Add a
/memoryslash command for viewing and managing persistent memories.List all memories:
View a specific memory:
Delete a memory:
Tell the assistant to remember something (natural language):
Alternatives Considered
read_memory/write_memorytools, but users cannot see or manage what's been saved without asking the model to list them — this lacks transparency.~/.iac-code/memory/, but this is not user-friendly.Related Area
CLI / Interactive Mode
Additional Context
The backend infrastructure (
MemoryManager,ReadMemoryTool,WriteMemoryTool, system prompt injection) already exists. This feature request is about adding a user-facing/memorycommand on top of it.