openclaw memory: Vector Memory and Semantic Search Guide

Vector Memory Overview

OpenClaw's Memory system gives the AI assistant long-term memory that extends beyond a single conversation's context window. It uses a vector database for storing and retrieving information with semantic search.

Enable the Memory System

openclaw memory enable

Configure vector storage:

{
  "memory": {
    "enabled": true,
    "provider": "local",
    "embeddingModel": "text-embedding-3-small",
    "embeddingProvider": "openai",
    "maxMemories": 10000,
    "searchResults": 5
  }
}

Manually Add Memories

# Add text memories
openclaw memory add "The company was founded in 2020, headquartered in Beijing"
openclaw memory add "John is the head of engineering, specializing in Python"

# Import from a file
openclaw memory import knowledge.txt
openclaw memory import faq.md --format markdown

# Bulk import from a directory
openclaw memory import ./docs/ --recursive

Search Memories

openclaw memory search "When was the company founded"

Search results (top 3):
  1. [0.92] The company was founded in 2020, headquartered in Beijing
     Source: manual  Added: 2026-03-15

  2. [0.78] The company raised Series A funding in 2021
     Source: docs/history.md  Added: 2026-03-15

  3. [0.65] The company currently has 200 employees
     Source: docs/about.md  Added: 2026-03-15

The number in brackets is the similarity score (0-1).

View Memory Statistics

openclaw memory stats

Memory Statistics:
  Total memories: 1,520
  Storage size: 25MB
  Embedding model: text-embedding-3-small
  Index status: up-to-date
  Last updated: 10 minutes ago

  By source:
    manual: 50
    docs/: 1,200
    conversations: 270

Manage Memories

# List recent memories
openclaw memory list --last 20

# Delete a specific memory
openclaw memory delete mem_abc123

# Clear all memories
openclaw memory clear --confirm

# Clear by source
openclaw memory clear --source "docs/"

Auto-Capture

Configure the AI to automatically extract and save important information from conversations:

{
  "memory": {
    "autoCapture": {
      "enabled": true,
      "threshold": 0.8,
      "categories": ["fact", "preference", "instruction"],
      "maxPerSession": 10
    }
  }
}

Using in AI Conversations

Enable memory tools for a channel:

{
  "channels": {
    "telegram-main": {
      "tools": ["memory_search", "memory_add"]
    }
  }
}

Example conversation:

User: Remember that I like coffee
AI: Got it, I've noted that you like coffee.

[A few days later]

User: What drink would you recommend for me?
AI: Based on my memory, you like coffee, so I'd recommend trying a latte or an americano.

Custom Embedding Models

Use a local embedding model:

{
  "memory": {
    "embeddingProvider": "ollama",
    "embeddingModel": "nomic-embed-text",
    "embeddingBaseUrl": "http://localhost:11434"
  }
}

Backup and Restore

# Back up memory data
openclaw memory export --output memory-backup.json

# Restore memory data
openclaw memory import --input memory-backup.json

Summary

The vector memory system gives OpenClaw's AI assistant long-term memory, allowing it to remember user preferences, company knowledge, and key information from past conversations. By combining auto-capture with manual knowledge import, you can build an AI assistant that gets smarter the more you use it.