[FEATURE] persistence of memory over chat sessions between chat sessions
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
Problem Statement
Claude Code CLI users lose 40-50% productivity between sessions due to complete memory loss. Every new session requires:
- Manually reconstructing context from previous work
- Re-searching for past decisions, discoveries, and solutions
- Repeating conversations and debugging steps
- Rebuilding understanding of project architecture and progress
This cognitive overhead is especially severe for long-running projects where institutional knowledge compounds. Users repeatedly ask themselves: "Did I solve this before? What was the decision?" with no way to search their history.
The core issue: Claude cannot automatically find and reference past work because there's no persistent, searchable memory system — users must manually tell Claude where information is, wasting time and breaking flow.
---
Proposed Solution
Proposed Solution
Implement a zero-configuration persistent memory system that:
- Automatically creates on first use with a simple user prompt (no installation, no plugins)
- Uses open-source tools: Fossil SCM + SQLite3 + FTS5 (all available on standard systems)
- Stores locally on user's machine (no cloud dependency, no privacy concerns)
- Enables automatic search: Claude searches user's history without being asked "where is X?"
- Persists across sessions: Accumulates valuable context over weeks/months of work
User experience:
- User runs:
claude --init-memoryor similar - System creates Fossil repo + SQLite FTS5 index on machine
- Claude automatically queries this memory when context is needed
- No configuration, no external services, no maintenance
---
Alternative Solutions
_No response_
Priority
Critical - Blocking my work
Feature Category
CLI commands and flags
Use Case Example
_No response_
Additional Context
_No response_
This issue has 4 comments on GitHub. Read the full discussion on GitHub ↗