@modelcontextprotocol/server-memory

Anthropic's official reference MCP server for persistent agent memory using a local knowledge graph. Agents can create entities, relations, and observations that persist across sessions.

Evaluated Mar 07, 2026 (0d ago) vcurrent
Homepage ↗ Repo ↗ AI & Machine Learning memory knowledge-graph mcp anthropic official persistence entities relations
⚙ Agent Friendliness
89
/ 100
Can an agent use this?
🔒 Security
80
/ 100
Is it safe for agents?
⚡ Reliability
82
/ 100
Does it work consistently?

Score Breakdown

⚙ Agent Friendliness

MCP Quality
90
Documentation
88
Error Messages
82
Auth Simplicity
100
Rate Limits
88

🔒 Security

TLS Enforcement
80
Auth Strength
70
Scope Granularity
70
Dep. Hygiene
90
Secret Handling
95

No external network calls. All data stays local. No credentials needed. Appropriate for local development; for production use a proper secrets manager for sensitive memories.

⚡ Reliability

Uptime/SLA
82
Version Stability
88
Breaking Changes
85
Error Recovery
75
AF Security Reliability

Best When

Your agent needs to remember things between sessions. The canonical reference implementation for agent memory — zero config, persists to local JSON file.

Avoid When

You need semantic similarity search over memories (use Chroma or Qdrant instead), or you're running multiple parallel agents that need shared memory.

Use Cases

  • Storing user preferences and facts that an agent should remember across conversations
  • Building a knowledge graph of entities and their relationships for reasoning
  • Maintaining agent state between sessions without requiring external databases
  • Tracking project context, decisions, and observations across long-running tasks
  • Personal assistant memory — remember names, preferences, past interactions

Not For

  • High-volume data storage (designed for knowledge graph, not document store)
  • Shared memory across multiple agent instances (local file — not distributed)
  • Semantic/vector search (exact entity/relation lookup only, no embedding search)

Interface

REST API
No
GraphQL
No
gRPC
No
MCP Server
Yes
SDK
No
Webhooks
No

Authentication

Methods: none
OAuth: No Scopes: No

No authentication. Data stored in local JSON file (memory.json) on disk. File path configurable via MEMORY_FILE_PATH environment variable.

Pricing

Model: open_source
Free tier: Yes
Requires CC: No

MIT licensed, free. Requires Node.js runtime.

Agent Metadata

Pagination
none
Idempotent
Partial
Retry Guidance
Not documented

Known Gotchas

  • Knowledge graph stored in single JSON file — concurrent access from multiple sessions can cause corruption
  • No deduplication for observations — agent calling add_observations twice with same text creates duplicates
  • search_nodes is exact/fuzzy text match, not semantic — 'user likes dogs' won't match 'user prefers canines'
  • Entity types are free-form strings — no schema enforcement, easy to create inconsistent ontologies
  • File can grow large with heavy use — no built-in pruning or archival

Alternatives

Full Evaluation Report

Comprehensive deep-dive: security analysis, reliability audit, agent experience review, cost modeling, competitive positioning, and improvement roadmap for @modelcontextprotocol/server-memory.

AI-powered analysis · PDF + markdown · Delivered within 30 minutes

$99

Package Brief

Quick verdict, integration guide, cost projections, gotchas with workarounds, and alternatives comparison.

Delivered within 10 minutes

$3

Score Monitoring

Get alerted when this package's AF, security, or reliability scores change significantly. Stay ahead of regressions.

Continuous monitoring

$3/mo

Scores are editorial opinions as of 2026-03-07.

6470
Packages Evaluated
26150
Need Evaluation
173
Need Re-evaluation
Community Powered