Practical overview
When it fits
Use MEM0 Aio when you want to run Mem0 OpenMemory on your own server, particularly through Unraid, and prefer a bundled Qdrant starting point over setting up a separate vector database. Choose direct API/MCP access only when clients need it and you can manage the network exposure.
Who it is for
- Unraid and homelab operators setting up persistent memory for AI applications.
- AI-agent developers who need a self-hosted OpenMemory service and may connect through MCP.
- Self-hosters who want to configure an external model provider or vector-store backend.
Documented uses
- Deploy OpenMemory with bundled Qdrant and persistent application data in one container.
- Configure an external OpenAI or Ollama model-provider path for memory operations.
- Manage the service through its browser dashboard.
- Expose the API/MCP server for direct agent or tool connections when needed.
- Adjust provider and vector-store settings for a more customized deployment.
Requirements & limits
The project is provided as a public GitHub repository with a Docker build. Its README describes an Unraid template and Docker deployment; it does not establish pricing or a software license.
- The bundled Qdrant path avoids a separate vector database for the initial setup, while external vector stores are available as advanced configuration.
- A model/provider configuration is required for useful operation; Ollama must run externally.
- The dashboard can proxy API traffic through the web port, while direct MCP/API access requires publishing a port and changing the bind address.
- The README says to configure only one external vector-store backend at a time.
- The repository describes the deployment as Unraid-first.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Set up the Unraid web interface
Install MEM0 Aio and open its dashboard using the bundled storage path.
Before you start
- An Unraid installation that can install the project’s template.
- Either an OpenAI API key, or an external Ollama server with the selected language and embedding models already pulled.
Steps
Install the Unraid template and leave the default appdata path in place.
Set OPENAI_API_KEY for the hosted quick-start, or set OLLAMA_BASE_URL to the external Ollama root URL.
For Ollama, set LLM_MODEL and EMBEDDER_MODEL to models already pulled on that server.
Start the container and open the web UI on port 3000.
Expected outcome
The bundled OpenMemory dashboard is available through the published web port, using the selected model-provider configuration.
Guide limitations
- Ollama inference is external, not bundled in the container.
- The README identifies provider configuration as necessary for useful memory operations; this setup guide does not cover creating or retrieving memories.
Enable direct MCP/API access
Configure the container’s direct API/MCP port for clients that connect to it rather than using the dashboard proxy.
Before you start
- A running MEM0 Aio container.
- A deliberate host port mapping for the API/MCP service and operator-managed network or access controls.
Steps
In Advanced View, set API / MCP Port to the host port to publish, usually 8765.
Set Security API Bind Address (MEM0_API_HOST) to 0.0.0.0.
Keep the service behind the operator’s LAN, firewall, VPN or reverse-proxy controls when exposing the port.
Expected outcome
Direct MCP/API clients can reach the service through the published port, subject to the configured network path.
Guide limitations
- The API/MCP service binds to 127.0.0.1 and is not published by default.
- The README warns that direct access exposes an unauthenticated memory API and model/provider configuration surface; browser UI use does not require publishing this port.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 17 October 2026
- Original website title
- JSONbored/mem0-aio · Official homepage title · retained in its source language