MCP Memory Service
Open-source persistent memory backend for AI agents that stores and retrieves context across sessions using hybrid BM25 + vector search, with optional Cloudflare sync for multi-device and multi-agent memory sharing.
Score Breakdown
⚙ Agent Friendliness
🔒 Security
Agent memory/persistence MCP. Stores potentially sensitive agent context. Local storage default — no network auth. For shared deployments, add auth layer. Encrypt sensitive memory payloads.
⚡ Reliability
Best When
You have long-running AI agent workflows or multi-session projects where context loss between sessions is a significant friction point, and you want local-first storage with optional cloud sync.
Avoid When
You need a fully managed cloud memory service with zero infrastructure setup, or you are on Python 3.13 and cannot downgrade. Also avoid if macOS system Python is your only option (requires Homebrew Python for sqlite extensions).
Use Cases
- • Persisting project decisions and architectural context across Claude Code sessions without re-explaining
- • Multi-agent pipelines where multiple AI agents share a common memory store with typed causal relationships
- • Reducing token overhead by externalizing long-term context from the AI's context window
- • Team collaboration scenarios where AI memory is synchronized across devices via Cloudflare Workers KV
- • Ingesting and semantically searching across PDF, TXT, MD, and JSON knowledge documents
Not For
- • Users on Python 3.13 (sqlite-vec compatibility issue; use Python 3.12)
- • Simple single-session use cases where in-context memory is sufficient
- • Teams needing a fully managed SaaS memory solution (requires self-hosting)
Interface
Authentication
Three modes: anonymous access (MCP_ALLOW_ANONYMOUS_ACCESS=true for local dev), API key (MCP_API_KEY env var for production), or OAuth 2.1 (MCP_OAUTH_ENABLED=true for team collaboration). HTTP dashboard requires auth by default. Local MCP usage can run without any auth.
Pricing
Apache 2.0 open source. Cloudflare sync feature may incur costs beyond free tier KV usage for high-volume memory stores.
Agent Metadata
Known Gotchas
- ⚠ Python 3.13 incompatible due to sqlite-vec missing pre-built wheels — use Python 3.12
- ⚠ macOS system Python lacks sqlite extension support — must use Homebrew Python
- ⚠ Cloudflare hybrid backend requires Cloudflare account setup before first use
- ⚠ MCP_MAX_RESPONSE_CHARS must be tuned or large memory retrievals can overflow agent context
- ⚠ Windows Claude Code users need a /session-start workaround for SessionStart hooks
- ⚠ X-Agent-ID header auto-tagging means multi-agent setups need careful agent identity management
Alternatives
Full Evaluation Report
Comprehensive deep-dive: security analysis, reliability audit, agent experience review, cost modeling, competitive positioning, and improvement roadmap for MCP Memory Service.
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 2026-03-07.