AnalyticDB:OpenClaw long-term memory solution based on Mem0

Step 1: Provision an OpenClaw instance (optional)

1. You can purchase a server instance as follows:

   • Go to the Simple Application Server purchase page.

   • For the region, select the same region as your AnalyticDB for MySQL cluster. This enables efficient internal network communication.

   • For the image, in the Application Image section, search for and select OpenClaw .

   • For the plan, to ensure stable operation, choose General-purpose 4 vCPUs and 16 GB RAM or higher.

   • Select your subscription duration and quantity as needed. Then, click Buy Now .

2. You can initialize the OpenClaw application as follows:

   • After the instance status changes to Running , click the instance ID to navigate to the Server Overview page.

   • Switch to the Application Details tab. Then, follow the on-screen instructions to complete Port Release and Configure OpenClaw .

   • In the Application Details > Model Configuration module, configure your large language model.

3. You can enable internal network interconnection as follows:

   • In the navigation pane on the left, select Internal Network Interconnection .

   • Click the Internal Network Interconnection button. This establishes an internal network connection between your Simple Application Server and other Alibaba Cloud services (such as AnalyticDB for MySQL) in the same region. This connection is essential for low-latency, secure access.

Step 2: Install the plugin and rebuild dependencies

At the top of the instance details page, click Remote Connection to log on to the server’s command-line terminal directly through your browser.

• Check your Python version. The plugin requires Python 3.8+ :

  # Check Python version
  python3 --version

• You can install Python 3.9 (optional). If your current version is below 3.8, run the following commands:

  # Install dependencies
  sudo yum install gcc openssl-devel bzip2-devel libffi-devel zlib-devel --disablerepo=docker-ce-stable
  
  # Download and install Python 3.9
  cd /tmp
  curl -O https://www.python.org/ftp/python/3.9.18/Python-3.9.18.tgz
  tar xzf Python-3.9.18.tgz
  cd Python-3.9.18
  ./configure --enable-optimizations
  sudo make altinstall
  
  # Verify installation
  python3.9 --version
  
  # Create a symbolic link
  sudo ln -sf /usr/local/bin/python3.9 /usr/bin/python3
  
  # Verify again
  python3 --version

• Install the plugin:

  openclaw plugins install @adbm-ai/openclaw-mem0

• Recompile better-sqlite3. Remove and reinstall better-sqlite3 compiled for Python 3.9:

  cd ~/.openclaw/extensions/openclaw-mem0
  rm -rf node_modules/better-sqlite3
  npm install better-sqlite3 --build-from-source

Step 3: Configure openclaw.json

This is a critical step. In the server’s command-line terminal, open the ~/.openclaw/openclaw.json file. Locate (or add) the "openclaw-mem0" configuration item, and fill in the following information based on your environment.

• Open the configuration file.

  Edit the ~/.openclaw/openclaw.json file. After installation, you will see the openclaw-mem0 field under plugins and entries.

  

• Configuration reference.

  Parameter	Description
  mode	Set to "open-source" to use self-hosted Mem0.
  userId	User identity for memory sharing and tenant isolation. Assign a unique userId to each OpenClaw instance.
  llm.provider	Fixed value: dashscope (Alibaba Cloud Model Studio).
  llm.config.apiKey	Use the API key for Alibaba Cloud Model Studio.
  llm.config.baseURL	https://dashscope-intl.aliyuncs.com/compatible-mode/v1.
  llm.config.model	Use a model supported by Alibaba Cloud Model Studio. We recommend qwen3.5-plus.
  embedder.provider	Fixed value: dashscope (Alibaba Cloud Model Studio).
  embedder.config.apiKey	API key from Alibaba Cloud Model Studio.
  embedder.config.url	https://dashscope-intl.aliyuncs.com/compatible-mode/v1.
  embedder.config.model	We recommend text-embedding-v4.
  embedder.config.embeddingDims	We recommend 1536 dimensions.
  vectorStore.provider	Fixed value: aliyun-adb-mysql.
  vectorStore.config.host	An internal network connection uses the VPC address of AnalyticDB for MySQL, and a public network connection uses the public network address of AnalyticDB for MySQL.
  vectorStore.config.port	Database port: 3306.
  vectorStore.config.database	openclaw_adb. Database name. If it does not exist or is unspecified, it will be created automatically. Do not modify this value.
  vectorStore.config.user	Username of the AnalyticDB for MySQL account. The account must have read and write permission.
  vectorStore.config.password	Password for the account.
  vectorStore.config.collectionName	openclaw_mem0_adb. Table name for memory storage.

• Configuration example.

  "openclaw-mem0": {
    "enabled": true,
    "config": {
      "mode": "open-source",
      "userId": "alice",
      "oss": {
        "llm": {
          "provider": "dashscope",
          "config": {
            "apiKey": "sk-2fa2a****cef0",
            "baseURL": "https://dashscope-intl.aliyuncs.com/compatible-mode/v1",
            "model": "qwen3.5-plus"
          }
        },
        "embedder": {
          "provider": "dashscope",
          "config": {
            "model": "text-embedding-v4",
            "embeddingDims": 1536,
            "apiKey": "sk-2fa2a****cef0",
            "url": "https://dashscope-intl.aliyuncs.com/compatible-mode/v1"
          }
        },
        "vectorStore": {
          "provider": "aliyun-adb-mysql",
          "config": {
            "host": "amv-0jl1****41.ads.aliyuncs.com",
            "port": 3306,
            "database": "openclaw_adb",
            "user": "your-user-name",
            "password": "your-password",
            "collectionName": "openclaw_mem0_adb"
          }
        }
      }
    }
  }

• Restart the service.

  openclaw gateway restart

  After configuration, you will see that the openclaw-mem0 plugin has loaded successfully.

Basic usage

• Start a conversation.

  openclaw tui

• Example 1: Add memory manually.

  Remember that I purchased an OpenClaw server from Alibaba Cloud. It expires on April 16, 2026.

• Example 2: Retrieve memory.

  What should I pay attention to recently?

Command-line interface (CLI) operations

• Run openclaw mem0 search "openclaw" to view memories.

  openclaw mem0 search "openclaw"

• Supported CLI commands include the following:

  # Search all memories (long-term + session)
  openclaw mem0 search "what languages does the user know"
  
  # Search only long-term memories
  openclaw mem0 search "what languages does the user know" --scope long-term
  
  # Search only session/short-term memories
  openclaw mem0 search "what languages does the user know" --scope session
  
  # Stats
  openclaw mem0 stats
  
  # Search a specific agent's memories
  openclaw mem0 search "user preferences" --agent researcher
  
  # Stats for a specific agent
  openclaw mem0 stats --agent researcher

Multi-agent long-term memory isolation

In OpenClaw’s multi-agent scenarios, the Mem0 + AnalyticDB for MySQL solution supports isolated long-term memory for multiple agents.

• Step 1: Create an agent.

  # Add writer agent
  openclaw agents add writer --workspace ~/.openclaw/workspace-writer --model bailian/qwen3.5-plus --non-interactive
  
  # Restart Gateway
  openclaw gateway restart
  
  # Verify agent list
  openclaw agents list

• Step 2: Interact with the agent.

  Run openclaw tui. Switch to the writer agent and start a conversation.

  

• Step 3: Search a specific agent’s memories.

  openclaw mem0 search "science fiction novel recommendations" --agent writer

• Step 4: Verify memory isolation.

  You can create a new agent reader.

  # Add reader agent
  openclaw agents add reader --workspace ~/.openclaw/workspace-reader --model bailian/qwen3.5-plus --non-interactive
  
  # Search reader's memories (should be empty)
  openclaw mem0 search "science fiction novel recommendations" --agent reader

  You can observe that memories are isolated across different agents.