Skip to main content

/memory:compact

Compress current session working memory into structured text optimized for session recovery and persistent storage.

Overview

The /memory:compact command compresses the current session's working memory into structured text, extracting critical information and saving it to persistent storage via MCP core_memory tool.

Parameters:

  • --description="...": Custom session description
  • --tags=<tag1,tag2>: Add custom tags
  • --force: Override existing memory without confirmation

Execution Flow:

  1. Session Analysis → 2. Structure Extraction → 3. Text Generation → 4. MCP Import

Features

  • Session Compression - Extracts key information from working memory
  • Structured Format - Organizes content for easy recovery
  • Critical State Capture - Preserves objectives, plans, and decisions
  • Tag Support - Add custom tags for organization
  • Persistent Storage - Saves via MCP core_memory tool
  • Session Recovery - Enables resuming from compacted state

Usage

# Compact current session
/memory:compact

# With custom description
/memory:compact --description="User authentication implementation"

# With tags
/memory:compact --tags=auth,security,api

# Force overwrite
/memory:compact --force

Execution Flow

Phase 1: Session Analysis

Analyze current session state to extract:

  • Session objectives and goals
  • Implementation plan
  • Files modified
  • Key decisions made
  • Constraints and requirements
  • Current state

Phase 2: Structure Extraction

const sessionAnalysis = {
objective: extract_objective(session),
plan: extract_plan(session),
files: extract_files(session),
decisions: extract_decisions(session),
constraints: extract_constraints(session),
state: extract_state(session),
notes: extract_notes(session)
};

Phase 3: Structured Text Generation

# Session: {session_id}

## Objective
{objective}

## Implementation Plan
{plan}

## Files Modified
{files}

## Key Decisions
{decisions}

## Constraints
{constraints}

## Current State
{state}

## Notes
{notes}

Phase 4: MCP Import

mcp__ccw-tools__core_memory({
operation: "import",
text: structuredText
})

Output Format

{
"operation": "import",
"id": "CMEM-YYYYMMDD-HHMMSS",
"message": "Created memory: CMEM-YYYYMMDD-HHMMSS"
}

Structured Text Template

# Session: {session_id}

**Date**: {timestamp}
**Description**: {custom_description or auto-generated}
**Tags**: {tags}

---

## Objective

{session_objective}

---

## Implementation Plan

{implementation_plan}

---

## Files Modified

| File | Changes |
|------|---------|
| {file1} | {changes1} |
| {file2} | {changes2} |

---

## Key Decisions

1. {decision1}
2. {decision2}
3. {decision3}

---

## Constraints

- {constraint1}
- {constraint2}

---

## Current State

{current_state}

---

## Notes

{additional_notes}

Examples

Basic Usage

# Compact current session
/memory:compact

# Output:
# Analyzing session...
# Extracting key information...
# Generating structured text...
# Importing to core memory...
# ✅ Memory compacted: CMEM-20250203-143022

With Custom Description

# Compact with description
/memory:compact --description="OAuth2 implementation"

# Description is saved with the memory

With Tags

# Compact with tags
/memory:compact --tags=oauth,authentication,security

# Tags help with organization and retrieval

Recovery

To recover a compacted session:

# List available memories
mcp__ccw-tools__core_memory({ operation: "list" })

# Export specific memory
mcp__ccw-tools__core_memory({ operation: "export", id: "CMEM-YYYYMMDD-HHMMSS" })

# Search for memories
mcp__ccw-tools__core_memory({ operation: "search", query: "oauth" })

Use Cases

  1. Session Handoff - Preserve context for later continuation
  2. Knowledge Base - Store insights and decisions for reference
  3. Team Sharing - Share session state with team members
  4. Documentation - Generate structured records of work sessions
  5. Recovery - Restore session state after interruption
  • /memory:load - Load project context into memory
  • /memory:update-full - Update all CLAUDE.md files
  • /memory:update-related - Update changed CLAUDE.md files

Notes

  • Persistent storage - Saved via MCP core_memory tool
  • Structured format - Optimized for readability and parsing
  • Automatic ID generation - Format: CMEM-YYYYMMDD-HHMMSS
  • Force option - Override existing memory without confirmation
  • Tag support - Add custom tags for organization and search
  • Custom description - Add meaningful description for easy identification