Consider enabling memory by default for project-based workflows

Preflight Checklist

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

Problem Statement

Context

I'm using Claude Code for ongoing Flutter development on a multi-month project. The current default relies on conversation compaction/summarization when approaching context limits.

Problem

With the current defaults:

• Token cost: Summaries are regenerated each time context limits are approached
• Performance: Summary generation adds latency to the session
• Context loss: Technical patterns and architecture decisions get compressed and lose nuance with each summarization cycle
• Inefficiency: Same patterns are re-summarized repeatedly across sessions

Observation

After manually enabling "Generate memory from chat history" in settings:

• Persistent context: Technical knowledge loads once per session instead of being repeatedly summarized
• Lower cost: Memory replaces expensive repeated summarization
• Better preservation: Critical patterns (like our frame type system, immutable update patterns, navigation conventions) remain intact
• Faster sessions: No summary generation overhead

Comparison

Cursor and similar dev-focused AI tools default to persistent memory/context for ongoing projects. This makes sense because:

• Development projects span weeks/months
• Architecture and patterns matter deeply
• Context builds over time
• The common case is ongoing work, not one-off tasks

Proposed Solution

Suggestion

Consider one of:

1. Default to memory-enabled for project-based work (detected by presence of .git, package.json, etc.)
2. Prompt users on first session to enable memory for projects
3. Add a setup wizard that asks about use case and recommends memory for development workflows

Benefits

• ✅ Lower token costs for long-running projects
• ✅ Better user experience (faster, more consistent context)
• ✅ Aligns defaults with the most common use case (ongoing development)
• ✅ Matches user expectations from other AI coding tools

Current Workaround

Manually enabling "Generate memory from chat history" works great, but requires users to discover this optimization themselves.

---

Note: This isn't a bug — the current behavior is fine for general use. This is purely a UX suggestion to optimize defaults for the most common development workflow.

And yes, Claude wrote all of this after summarizing our chat about this issue. :)

Alternative Solutions

Priority

Low - Nice to have

Feature Category

CLI commands and flags

Use Case Example

_No response_

Additional Context

_No response_