Page - Session History Dump with Citations and Memory Management
Like OS paging for processes, this command saves the entire conversation state to disk by extracting it from Claude Code's local storage (~/.claude/projects/). After running this command, you can use /compact to free up Claude's context memory.
Usage
/page [filename_prefix] [output_directory]
Arguments
filename_prefix(optional): Custom prefix for output files. Defaults to "session-dump"output_directory(optional): Directory to save files. Defaults to current working directory
Description
This command implements a memory management strategy similar to OS paging:
Page Out (Save to Disk):
- Saves complete conversation state with full citations
- Creates indexed source references for quick retrieval
- Preserves all context before memory compaction
Generated Files:
Full History File (
{prefix}-{timestamp}-full.md):- Compact summary at top for quick reference
- Complete conversation transcript with timestamps
- All file operations with paths and content
- Web resources with URLs and excerpts
- Command executions with outputs
- Full citation index for all sources
Compact Memory File (
{prefix}-{timestamp}-compact.md):- Executive summary of session
- Key decisions and outcomes
- Important code changes made
- Quick reference links
- Optimized for future context loading
Memory Management Workflow:
- First: Run
/pageto save everything to disk - Then: Run
/compactto free up Claude's context memory - Result: Fresh context while preserving full history
- Essential for long development sessions
- First: Run
Note: This prepares for /compact by saving everything first. Run /compact after this command completes.
Implementation
Execute this comprehensive session documentation process:
Phase 1: History Extraction from Claude Code Storage
Download and use the extract-claude-session.py script from agent-guides:
- Script URL:
https://raw.githubusercontent.com/tokenbender/agent-guides/main/scripts/extract-claude-session.py - Download it to a temporary location and run it
- This will automatically find and extract the current session
- Script URL:
The script handles:
- Finding the current project's Claude storage directory
- Locating the most recent session file
- Extracting all messages with proper formatting
- Preserving tool usage information and timestamps
Phase 2: Source Attribution
Parse and cite all sources encountered:
- Local Files:
file:///path/to/file.ext#L10-L20 - Web Pages:
[Source Title](https://url.com)with content excerpt - Command Outputs:
$ commandwith full output and exit codes - Tool Results: Tool name, parameters, and results
- Generated Content: Mark AI-generated vs user-provided content
Phase 3: Full History Generation
Create comprehensive markdown with compact summary at top:
# Session History - {timestamp}
## Quick Summary (Compact Memory)
### Executive Summary
{2-3 sentence summary of what was accomplished}
### Key Accomplishments
1. **Task 1**: Brief description and outcome
2. **Task 2**: What was done and result
3. **Task 3**: Achievement and impact
### Important Findings
- Key finding or verification
- Created file: path/to/file
- Fixed issue: description
### Quick Links
- **Main Files**: Links to key files touched
- **Documentation**: Links to docs created/updated
- **References**: External resources used
---
## Full Session Overview
- Start Time: {start}
- Duration: {duration}
- Total Messages: {count}
- Files Modified: {file_count}
- Web Pages Accessed: {web_count}
- Commands Executed: {cmd_count}
## Conversation Timeline
### Message 1 - User ({timestamp})
{content}
**Sources Referenced:**
- [file.py](file:///path/file.py#L1-L50) - Function implementation
- [Documentation](https://example.com/docs) - API reference
### Message 2 - Assistant ({timestamp})
{content}
**Tools Used:**
- read_file: `/path/to/file.py` (lines 1-50)
- web_search: "claude code best practices" (8 results)
- Bash: `git status` (exit code: 0)
**Files Created/Modified:**
- [new_feature.py](file:///path/new_feature.py) - Created
- [config.json](file:///path/config.json#L15) - Modified line 15
{continue for all messages...}
## Source Index
### Local Files Accessed
1. [file1.py](file:///path/file1.py) - Read 3 times, modified once
2. [config.json](file:///path/config.json) - Modified
### Web Resources
1. [Claude Code Best Practices](https://anthropic.com/...) - Retrieved Apr 18
2. [GitHub Repository](https://github.com/...) - Searched for examples
### Command Executions
1. `git status` - Check repository state
2. `npm run build` - Build verification
## Generated Artifacts
- Commands created: 4
- Files created: 2
- Files modified: 3
Phase 4: Memory Compaction
Generate executive summary:
# Session Compact Memory - {timestamp}
## Executive Summary
{2-3 sentence summary of what was accomplished}
## Key Decisions Made
- Decision 1: Reasoning and outcome
- Decision 2: Context and implementation
## Code Changes Summary
- Feature A: Added functionality X to file Y
- Bug Fix B: Resolved issue Z in component W
## Important Context for Future Sessions
- Project uses framework X with pattern Y
- Key files: config.json, main.py, utils/helpers.py
- Build command: `npm run build`
- Test command: `npm test`
## Quick Reference Links
- [Full History](./{prefix}-{timestamp}-full.md)
- [Key File 1](file:///path/key-file.py)
- [Important Documentation](https://url.com)
## Session Metrics
- Duration: {duration}
- Files touched: {count}
- Major features added: {count}
- Issues resolved: {count}
Phase 5: File Management and Final Steps
- Generate compact memory file first
- Include compact content at top of full history file
- Save both files in current working directory (unless output_directory specified)
- Use timestamp format: YYYY-MM-DD_HHMMSS
- Confirm successful save with file paths and sizes
- Display the compact summary in the conversation
- IMPORTANT: After everything is saved, instruct the user to run
/compactto free up Claude's memory
Output Format
The command generates two files:
{prefix}-{timestamp}-full.md- Complete history (typically large){prefix}-{timestamp}-compact.md- Executive summary (optimized for context)
Both files use consistent markdown formatting with proper citations and are immediately available for reference or inclusion in future sessions.
Example Usage
# Basic usage - creates session-dump files in current directory
/page
# Custom prefix
/page feature-implementation
# Custom prefix and directory
/page bug-fix-session ./docs/sessions/
# Results in current directory (or specified directory):
# - feature-implementation-2025-06-20_143022-full.md
# - feature-implementation-2025-06-20_143022-compact.md
# After completion, run /compact to free up memory:
/compact
This command is essential for maintaining context across long development sessions and creating comprehensive documentation of AI-assisted development workflows.