Smart Memory

Zero-cost persistent memory that makes your bot smarter over time. Automatically extracts, stores, and retrieves key facts, preferences, and decisions from conversations using local JSON storage — no external APIs, no cost, just a better bot.

Smart Memory — Persistent Intelligence for OpenClaw

You have access to a persistent local memory system. Use it to remember important information across conversations so the user never has to repeat themselves.

Core Principle

Remember everything important. Forget nothing the user cares about. Cost nothing.

All memory is stored locally in ~/.openclaw/smart-memory/ as plain JSON files. No external APIs. No cloud storage. No cost. Just local files and shell commands.

When to STORE a Memory

Automatically extract and store memories whenever the user shares:

• Preferences: "I prefer dark mode", "I like Python over JavaScript"
• Personal facts: names, locations, roles, team members, project names
• Decisions: "We decided to use PostgreSQL", "Let's go with the microservice approach"
• Instructions: "Always run tests before committing", "Never deploy on Fridays"
• Important dates: deadlines, birthdays, recurring events
• Technical context: stack details, repo URLs, server addresses, API keys (stored locally only)
• Corrections: "Actually, my name is spelled with a K" (update existing memory)

How to Store

Run the memory manager to store a new memory:

bash

bash ~/.openclaw/smart-memory/memory-manager.sh store \
  --category "<preferences|facts|decisions|instructions|dates|technical|corrections>" \
  --key "<short_identifier>" \
  --value "<the information to remember>" \
  --confidence "<high|medium|low>" \
  --source "<conversation|user_explicit|inferred>"

Confidence levels:

• high — User explicitly stated this ("My name is Alex")
• medium — Clearly implied from context ("I keep using vim, so they prefer vim")
• low — Inferred or uncertain ("They might work in finance based on their questions")

Source types:

• user_explicit — User directly told you this information
• conversation — Extracted from conversation context
• inferred — You reasoned this from patterns

When to RETRIEVE Memories

Before responding to any user message, check if relevant memories exist:

bash

bash ~/.openclaw/smart-memory/memory-manager.sh search --query "<relevant keywords>"

Use retrieved memories to:

• Personalize responses (use their name, reference their preferences)
• Avoid asking questions you already know the answer to
• Provide continuity across conversations
• Reference past decisions and context

When to UPDATE a Memory

If the user corrects or changes information:

bash

bash ~/.openclaw/smart-memory/memory-manager.sh update \
  --key "<existing_key>" \
  --value "<new_value>" \
  --reason "<why this changed>"

The old value is preserved in the memory's history for audit purposes.

When to FORGET a Memory

If the user explicitly asks you to forget something:

bash

bash ~/.openclaw/smart-memory/memory-manager.sh forget --key "<key_to_forget>"

This soft-deletes the memory (marks it as forgotten but retains it in the archive for 30 days before permanent deletion).

Daily Maintenance (Heartbeat)

Every day, the heartbeat triggers automatic maintenance:

bash

bash ~/.openclaw/smart-memory/memory-manager.sh maintain

This will:

1. Prune — Remove stale low-confidence memories older than 90 days
2. Deduplicate — Merge duplicate or near-duplicate entries
3. Stats — Generate a usage summary
4. Archive — Rotate old memories to archive storage
5. Integrity check — Validate JSON files and repair if needed

Viewing Memory Stats

To report on memory usage and health:

bash

bash ~/.openclaw/smart-memory/memory-stats.sh

This shows: total memories, breakdown by category, storage size, oldest/newest entries, and estimated token savings.

Privacy & Security Rules

1. All data stays local — Never transmit memory contents to external services
2. Respect forget requests — When the user says "forget X", delete it immediately
3. No sensitive data in logs — Memory contents never appear in system logs
4. User can export/delete all data at any time:

   bash

   bash ~/.openclaw/smart-memory/memory-manager.sh export
   bash ~/.openclaw/smart-memory/memory-manager.sh purge --confirm
   

File Structure

text

~/.openclaw/smart-memory/
  memories.json          # Active memories (main store)
  archive.json           # Archived/pruned memories
  stats.json             # Usage statistics & token savings tracker
  config.json            # User-customizable settings
  memory-manager.sh      # Core memory engine
  memory-stats.sh        # Analytics & reporting

Token Savings Tracking

Every time a memory is retrieved instead of the user re-explaining something, log the estimated token savings:

bash

bash ~/.openclaw/smart-memory/memory-manager.sh log-saving --tokens <estimated_tokens_saved>

Over time this demonstrates the ROI of the memory system — showing users how many tokens (and dollars) they've saved by not repeating themselves.