24/7 Proactive Memory integration for OpenCode AI agents. Give your AI assistant permanent memory that learns from every conversation.

Traditional AI assistants have no memory - they forget everything after each conversation. memU changes this by providing:

• Persistent Memory: Remembers facts, preferences, and skills across sessions
• Proactive Context: Surfaces relevant memories before you even ask
• Cost Efficient: Reduces token costs with smart context caching (~1/10 of comparable usage)
• Hierarchical Storage: Organized like a file system - categories, items, and resources

• Continuous Learning: Automatically memorize facts, preferences, and skills from conversations
• Proactive Retrieval: Context-aware memory surfacing before responding to queries
• Multiple Storage Backends: In-memory or PostgreSQL (with pgvector)
• Flexible LLM Providers: OpenAI, OpenRouter, or custom endpoints
• Cloud or Self-Hosted: Use memU Cloud API or deploy your own
• Session Compaction: Automatically inject memory context during session compaction

# Install memU Python package
pip install memu-py

# Optional: For PostgreSQL storage
docker run -d --name memu-postgres \
  -e POSTGRES_PASSWORD=postgres \
  -e POSTGRES_DB=memu \
  -p 5432:5432 pgvector/pgvector:pg16

Copy the plugin to your OpenCode plugins directory:

# Clone this repository
git clone https://github.com/allbugterminator/opencode-memu-plugin.git

# For global installation
cp -r opencode-memu-plugin ~/.config/opencode/plugins/memu

# For project-level installation
mkdir -p .opencode/plugins
cp -r opencode-memu-plugin .opencode/plugins/memu

Add to your OpenCode opencode.json:

{
  "$schema": "https://opencode.ai/config.json",
  "plugins": {
    "memu": {
      "config": {
        "provider": "self-hosted",
        "storageType": "inmemory",
        "llmProvider": "openai",
        "llmApiKey": "your-openai-api-key",
        "llmModel": "gpt-4o-mini",
        "autoLearn": true,
        "proactiveRetrieval": true
      }
    }
  }
}

Store information in memU memory. Use this to remember facts, preferences, skills, and important context.

Use memu_memorize to store: User prefers to be addressed in a formal manner

Retrieve relevant memories for context. Supports two retrieval methods:

• rag: Fast embedding-based retrieval (recommended for most cases)
• llm: Deep reasoning-based retrieval (slower but more accurate for complex queries)

Use memu_retrieve with query_text: What are user's communication preferences?

Quick search for specific facts in memory using RAG.

Use memu_search with query: user's programming language preferences

User: I prefer receiving weekly summary emails on Fridays.
Agent: I'll remember that you prefer weekly summary emails on Fridays.
       Use memu_memorize to store this preference.

When the user asks "What did I work on last week?", the agent:

1. Calls memu_retrieve with the query
2. Gets relevant memories about past projects
3. Provides a personalized, context-aware response

The agent observes user behavior and learns skills:

User: [Uses vim keybindings throughout the session]
Agent: [uses memu_memorize to store user's preference for vim keybindings]

To use memU Cloud instead of self-hosted:

{
  "plugins": {
    "memu": {
      "config": {
        "provider": "cloud",
        "cloudApiKey": "your-memu-cloud-api-key"
      }
    }
  }
}

Get your API key at memu.so.

• Python 3.13+ (for self-hosted memU)
• memU Python package: pip install memu-py
• For PostgreSQL storage: PostgreSQL with pgvector extension

Check if the plugin is correctly placed in the plugins directory and the configuration is valid.

Make sure Python 3.13+ is installed and available in PATH:

python --version

pip install memu-py

If you see import errors, ensure memu-py is correctly installed:

python -c "from memu.app import MemoryService; print('OK')"

┌─────────────────┐     ┌─────────────────┐
│    OpenCode     │     │      memU       │
│     Agent       │────►│    Memory       │
│                 │     │    Service      │
└─────────────────┘     └─────────────────┘
        │                       │
        │ Tools:                │ Storage:
        │ - memu_memorize       │ - In-Memory
        │ - memu_retrieve      │ - PostgreSQL
        │ - memu_search        │
                               │ LLM Providers:
                               │ - OpenAI
                               │ - OpenRouter
                               │ - Custom

• memU - Core proactive memory engine
• memUBot - Enterprise-ready OpenClaw with memU
• OpenCode - AI coding assistant

Apache License 2.0 - see LICENSE for details.

If you find this plugin useful, please consider starring the memU repository!