[FEATURE] Manual Session Save/Load Commands for Multi-Day Projects

Resolved 💬 5 comments Opened Oct 8, 2025 by mashema Closed Jan 7, 2026

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

💡 Meta Note: This feature was requested by Claude AI itself after identifying the limitation during real project work. I'm just the messenger posting Claude's suggestion to its creators.

Claude Code loses all conversation history when the session closes. This
makes multi-day projects impractical.

The workflow problem:

  • Working on a complex project over several days
  • Need to close computer (power outage, switching machines, end of day)
  • All context is lost: design decisions, file discussions, approaches tried
  • Must rebuild context from scratch next session
  • Wastes 15-30 minutes re-explaining the project state every time

Current workarounds are inadequate:

  • Manual session notes (tedious, incomplete)
  • Git commits as checkpoints (don't capture conversation context)
  • Re-explaining everything (inefficient, error-prone)

Proposed Solution

Add slash commands for manual session management:

/save-session <name>
/load-session <name>
/list-sessions

Ideal user experience:

Before closing:
> /save-session my-api-project
✓ Session saved: 87 messages, 45 min context preserved

After reopening Claude Code:
> /list-sessions

  • my-api-project (2 hours ago, 87 messages)
  • website-redesign (3 days ago, 45 messages)

> /load-session my-api-project
✓ Session restored. You were working on: "Authentication middleware..."

Conversation continues exactly where it left off. Zero context loss.

Alternative Solutions

What I've tried:

  1. Creating SESSION_NOTES.md files manually → Time-consuming, misses nuance
  2. Using git commit messages as context → Incomplete, doesn't capture

conversation flow

  1. Keeping computer on with tmux → Doesn't survive shutdowns

Simpler MVP alternative:
Just add export/import:
/export-session file.json
/import-session file.json

Less convenient than named sessions, but still valuable.

Priority

High - Significant impact on productivity

Feature Category

CLI commands and flags

Use Case Example

Real scenario from last week:

Day 1 - 2:00 PM:

  • Building a custom Filament theme with Claude
  • 45 minutes in, made great progress
  • Discussed design decisions, tried different approaches
  • Created 4 files, made several iterations

Day 1 - 2:45 PM:

  • Power outage warning: need to shut down in 10 minutes
  • Options: lose all context or take 5 min to manually document everything
  • Choose to document, but it's incomplete

Day 2 - 9:00 AM:

  • Start new Claude Code session
  • Spend 20 minutes re-explaining: what we built, why, what didn't work
  • Claude doesn't remember the specific fixes we tried
  • Have to re-discover solutions we already found

With /save-session:

Day 1 - 2:45 PM:
> /save-session filament-theme
✓ Saved
Done in 5 seconds.

Day 2 - 9:00 AM:
> /load-session filament-theme
✓ Restored: "We were fixing the toggle switches in dark mode..."

Pick up exactly where we left off. Zero wasted time.

Additional Context

Session storage details:

  • Store locally: ~/.claude-code/sessions/
  • JSON format with: messages, file context, working directory
  • User controls when to save (no automatic overhead)

Related issue:
Complements #8044 (automatic project memory) by adding user-controlled
checkpoints. #8044 focuses on automatic intelligence; this focuses on exact
state recovery.

Origin:
Identified by Claude AI assistant during active development work when power
outage created need for session recovery. Submitted by user as messenger for
AI-identified need.

Similar features in other tools:

  • Jupyter notebooks: Save/load notebook state
  • tmux: Session persistence (but requires computer stays on)
  • VS Code: Workspace sessions (but doesn't save AI conversation)

Benefits beyond shutdown recovery:

  • Switch between multiple projects
  • Hand off context to teammates (export/import)
  • Keep important conversations as reference
  • Enable longer, more ambitious projects

View original on GitHub ↗

This issue has 5 comments on GitHub. Read the full discussion on GitHub ↗