← Back to Mem0 MCP Self-hosted

Install Mem0 MCP Self-hosted

Pick your client, copy the command, done.

ai-tools
1

Add it to Claude Code

claude mcp add -e "MEM0_USER_ID=${MEM0_USER_ID}" mem0-mcp-selfhosted -- uvx --from git+https://github.com/elvismdev/mem0-mcp-selfhosted.git mem0-mcp-selfhosted
Required:MEM0_USER_ID+ 2 optional

Environment Variables

Set these before running Mem0 MCP Self-hosted.

VariableDescriptionRequired
MEM0_USER_IDUnique identifier for the userYes
MEM0_PROVIDERLLM provider for memory processing (e.g., ollama)No
MEM0_LLM_MODELModel to use for memory processingNo

Available Tools (3)

Once configured, Mem0 MCP Self-hosted gives your AI agent access to:

search_memoriesSearch for relevant context in memory.
query
add_memoryAdd new information to persistent memory.
data
update_memoryUpdate existing memory entries.
memory_iddata

Try It Out

After setup, try these prompts with your AI agent:

→Search my memories for TypeScript preferences.
→Remember that I prefer Hatch for Python packaging.
→Show me all entities in my knowledge graph.
→What do I usually do when I encounter a 404 error in this project?
Prerequisites & system requirements
  • An MCP-compatible client (Claude Code, Cursor, Windsurf, Claude Desktop, or Codex)
  • Python 3.8+ with pip installed
  • MEM0_USER_ID — Unique identifier for the user

Keep this setup from going cold

Save the docs, env vars, and workflow around Mem0 MCP Self-hosted in Conare so Claude Code, Codex, and Cursor remember it next time.

Remember this setup
← Mem0 MCP Self-hosted overview Browse ai-tools servers →