Feature Request: Persistent Project Memory to avoid repeating operational facts

Problem

Users have to repeatedly explain the same project-specific operational facts across sessions - things like build commands, folder structures, and deployment steps.

Real example: A user reports explaining their build process approximately 750 times because each session starts fresh. Their build command is simply cd cryptoking && npm run build - this never changes, yet Claude Code repeatedly runs the wrong command, breaks things, and requires re-explanation.

Current Behavior

• Each Claude Code session is stateless
• Context summaries carry over some information but lose operational details
• When a command fails, Claude makes assumptions and 'fixes' things instead of first reading existing documentation
• This wastes significant user time and causes real frustration

Requested Solution

A persistent, local project-level database/memory store that:

1. Stores operational facts - build commands, folder structures, common workflows, project-specific conventions
2. Automatically loads at session start - no user action required
3. Is project-scoped - per repository/working directory
4. Can be updated - by users directly or learned from corrections

Why This Matters

The information that gets lost is:

• Static (doesn't change)
• Project-specific (not general knowledge)
• Repeatedly needed (every session)
• Already documented (in README, config files, etc.)

A simple key-value store per project would solve this entirely. The current architecture of stateless sessions + context summaries clearly fails for operational knowledge retention.

Impact

This limitation causes:

• Wasted user time re-explaining basics
• Broken builds from wrong commands
• User frustration from repetitive corrections
• Reduced trust in the tool

Thank you for considering this enhancement.