mcp-mem0-docker

# MCP-Mem0-docker: Long-Term Memory for AI Agents ## Goals - Own your memories, Own your data! - Run locally on any computer with Docker and Ollama. ### Plus - Use any LLM provider, currently (OpenAI, OpenRouter, Ollama) - Provide a simple and effective way to manage long-term memory for AI agents using the MCP protocol. - Enhance your AI's capabilities with persistent memory storage. ## Features Based on [mem0.ai](https://mem0.ai/) Provides 4 tools for managing long-term memory: 1. **`memory_save`**: Store any information in long-term memory. 2. **`memory_get_all`**: Retrieve stored memories in contextual format. 3. **`memory_search`**: Find Memories using semantic search 4. **`memory_delete`**: Delete memories by searching an specific. ## Prerequisites - Ollama running locally. - Docker. - Any MCP client (VS Code, Claude Desktop, Cursor, Windsurf, etc.) ## Fast Start ### Pull models and embeddings to Ollama Make sure you have Ollama installed and running locally. You can pull the models and embeddings using the following commands: ````bash ollama pull qwen2.5:3b ollama pull nomic-embed-text ollama serve ```` ### Run the MCP Server and the database ```bash docker compose up -d ``` ### Configure the MCP Client For example, if you are using the MCP client in VS Code, you can create a configuration file like this: in your project root: `~/.mcp/config.json`: ```json { "mcpServers": { "mem0": { "transport": "sse", "url": "http://localhost:8050/sse" } } } ``` #### Windsurf Use `serverUrl` instead of `url` in your configuration: ## Development Best way to develop or tweak around is to go to Fast Start section and run the server/database with docker compose, then you can stop the MCP server and keep the database running, and run the server locally with uv: - Python 3.12+ - uv ### Using uv 1. Install uv: ```bash pip install uv uv venv .venv .venv\Scripts\activate ``` 2. Install dependencies: ```bash uv pip install -e . ``` 3. Create a `.env` file based on `.env.example`: ```bash cp .env.example .env ``` 4. Configure your environment variables in the `.env` file (see Configuration section) 5. Start coding and the server locally: ```bash uv run src/main.py ``` 6. Rebuild with docker compose no cache: ```bash docker compose up -d --build --no-cache ``` ### Configuration .env The following environment variables can be configured in your `.env` file: | Variable | Description | Default .env | Notes | |----------|-------------|--------------| ------| | `TRANSPORT` | Transport protocol (sse or stdio) | `sse` | Recommended to use `sse` for better performance | | `HOST` | Host to bind to when using SSE transport | `0.0.0.0` | This allows access from any IP address | | `PORT` | Port to listen on when using SSE transport | `8050` | Change if needed | | `LLM_PROVIDER` | LLM provider (openai, openrouter, or ollama) | `ollama` | Use `ollama` for local models | | `LLM_BASE_URL` | Base URL for the LLM API | `http://localhost:11434` | Default for Ollama | | `LLM_API_KEY` | API key for the LLM provider | Empty | Required for OpenAI and OpenRouter | | `LLM_MODEL` | LLM model to use | `qwen2.5:3b` | Change to your desired model | | `LLM_EMBEDDING_MODEL` | Embedding model to use | `nomic-embed-text` | Change to your desired embedding model | | `DATABASE_URL` | PostgreSQL connection | `postgresql://user:pass@host:port/db` | Change to your PostgreSQL connection string | #### Note for n8n users Use `host.docker.internal` instead of `localhost` so that the MCP server can be accessed from the n8n container. This is necessary because `localhost` in the n8n container refers to the container itself, not your host machine. The URL in the MCP node would be: <http://host.docker.internal:8050/sse>