modbender OpenClaw Mem0 Plugin
Mem0 integration for OpenClaw. Adds intelligent long-term memory to your agents, allowing them to remember user preferences, facts, and past conversations automatically.
When to use
- You want your agent to remember user details (name, job, preferences) across sessions
- You need "infinite context" by retrieving relevant past interactions
- You want to build a personalized assistant that learns over time
- You need both cloud (managed) and self-hosted (local) memory options
Setup
Platform Mode (Recommended)
- Get a free API key at mem0.ai
- Add to your OpenClaw config:
{
"plugins": {
"entries": {
"openclaw-mem0": {
"enabled": true,
"config": {
"mode": "platform",
"apiKey": "your-mem0-api-key",
"userId": "default-user"
}
}
}
}
}
Open-Source Mode (Self-Hosted)
Connect to your own Mem0 instance (requires mem0ai package installed):
{
"plugins": {
"entries": {
"openclaw-mem0": {
"enabled": true,
"config": {
"mode": "open-source",
"oss": {
"vectorStore": {
"provider": "chroma",
"config": {
"collectionName": "memories",
"path": "./chroma_db"
}
}
}
}
}
}
}
}
Usage
This plugin works automatically (Zero-Shot) but also provides manual tools.
Automatic Features
- Auto-Recall: Before every agent turn, it searches memory for relevant context and injects it into the system prompt.
- Auto-Capture: After every agent turn, it analyzes the conversation and stores key facts into memory.
Manual Tools
The agent can proactively call these tools:
| Tool | Description | Parameters |
|---|---|---|
memory_store |
Explicitly save a fact | text (string), longTerm (bool) |
memory_search |
Search memories | query (string), scope ("session"|"long-term") |
memory_get |
Get memory by ID | memoryId (string) |
memory_list |
List all memories | userId (string) |
memory_forget |
Delete a memory | memoryId (string) or query (string) |
Example
User: "I'm moving to Tokyo next month." Agent automatically captures this fact.
(Two weeks later) User: "What's a good restaurant for my farewell dinner?" Agent automatically recalls "User is moving to Tokyo" and suggests a restaurant in their current city.
Plugin structure
openclaw-mem0/
package.json # NPM package config (@xray2016/openclaw-mem0)
index.ts # Plugin implementation & tools
lib/ # Internal Mem0 client implementation
SKILL.md # This file
README.md # Detailed documentation
Author
Maintained by @xRay2016. Modified from the original Mem0 OpenClaw integration.