claude-code - 💡(How to fix) Fix [FEATURE] Memory system: nudge toward consolidation to prevent redundant file creation [6 comments, 4 participants]

Preflight Checklist

• I have searched existing requests and this feature hasn't been requested yet
• This is a single feature request (not multiple features)

Problem Statement

When using auto memory for the first time, it's easy to end up with multiple memory files when one would do. The system has 4 types (user, feedback, project, reference) which encourages splitting things up. For new users with very little to store, this means 3-4 files with overlapping info instead of 1 consolidated file.

In my session today helping a friend file his first PR, Claude created 3 separate memory files for about 11 lines of useful content. We then spent more tokens consolidating them into one file. Roughly $1-2 wasted on redundant memory writes that could've been one file from the start. The friction is silent. There's no nudge or warning telling new users they can put it all in one file. Power users figure this out, but everyone burns tokens learning it.

Proposed Solution

A simple consolidation nudge when memory files get created. A few options:

1. Heuristic check: if total memory content is under ~30 lines, suggest one file instead of multiple.
2. Doc note in the auto-memory section: "Start with one file. Split into separate files only when content grows past ~50 lines or you have genuinely distinct categories."
3. In-conversation hint: when Claude is about to create a 2nd memory file with little content in the 1st, suggest combining them.

Doesn't need to be enforced. Just a nudge so new users don't waste tokens unknowly.

Alternative Solutions

Manually editing the memory file outside the conversation works and is much cheaper than asking Claude to do it. But I only found this out by asking. New users won't know it's an option until they've already wasted tokens.

Priority

Low - Nice to have

Feature Category

Other

Use Case Example

New Claude Code user enables auto memory in their first long session. Claude saves preferences across 3 separate type-based files (~11 lines total content) plus a MEMORY.md index. User sees the token cost on /usage and asks why so many files. They spend more tokens consolidating to 1 file. A nudge at file # 2 ("you have very little in file # 1, consider combining") would have prevented the back-and-forth.

Additional Context

Every new user will likely create multiple memory files like I did, wasting tokens. A nudge or auto-consolidation suggestion saves every future user money: like Pro users on usage limits.

Related but distinct from existing memory issues that address where memory lives or how it scales. This one is about the entry point preventing waste before users hit those scaling questions?