mem0-mcp-selfhosted
Provides a self-hosted MCP server exposing mem0 memory-management tools (and optional Neo4j graph tools) for use with Claude Code, backed by Qdrant for vector search, Ollama for embeddings and optionally LLM, and optional Anthropic/OAT integration for LLM-based extraction.
Score Breakdown
⚙ Agent Friendliness
🔒 Security
README indicates tokens can be read from local Claude credentials and supports env-var overrides. It also mentions OAT refresh and defensive fallback. However, the provided content does not describe TLS enforcement for internal service URLs (e.g., Qdrant/Neo4j endpoints) nor offer details on how secrets are logged. Scope granularity for auth/authorization of tools is not discussed; access control likely relies on the MCP transport/session context.
⚡ Reliability
Best When
You want self-hosted AI memory with Claude Code using MCP/stdio, and you can run Qdrant + Ollama (and optionally Neo4j).
Avoid When
You need strict, centrally managed security/compliance controls without handling tokens/models locally; or you cannot provide/secure Anthropic or Ollama dependencies.
Use Cases
- • Persistent cross-session memory for coding assistants (Claude Code)
- • Semantic retrieval of previously learned facts and preferences
- • Optional knowledge graph storage and retrieval for entity relationships
- • Automated session-boundary memory injection and summarization via Claude Code hooks
Not For
- • Production deployments requiring a managed, fully hosted memory service
- • Teams needing a documented public REST/SDK API for direct programmatic use beyond MCP
- • Environments that cannot run supporting infrastructure (Qdrant and Ollama; Neo4j optional)
Interface
Authentication
Authentication is primarily delegated to Claude Code session credentials (OAT token). The server also supports using Anthropic API keys or OAT via env vars and includes proactive OAT refresh/defensive fallback on auth failures per README.
Pricing
Self-hosted; no direct SaaS pricing described in the provided content.
Agent Metadata
Known Gotchas
- ⚠ Graph tooling (Neo4j + graph LLM steps) may increase latency; hooks explicitly disable graph to stay within timeout budgets.
- ⚠ Some configuration defaults assume local services (Qdrant localhost:6333, Ollama localhost:11434).
- ⚠ If Ollama embedding model dimensions or Qdrant collection configuration mismatch (MEM0_EMBED_DIMS), operations may fail.
Alternatives
Full Evaluation Report
Comprehensive deep-dive: security analysis, reliability audit, agent experience review, cost modeling, competitive positioning, and improvement roadmap for mem0-mcp-selfhosted.
AI-powered analysis · PDF + markdown · Delivered within 30 minutes
Package Brief
Quick verdict, integration guide, cost projections, gotchas with workarounds, and alternatives comparison.
Delivered within 10 minutes
Score Monitoring
Get alerted when this package's AF, security, or reliability scores change significantly. Stay ahead of regressions.
Continuous monitoring
Scores are editorial opinions as of 2026-03-30.