MCP Server for Persistent Memory — Model Context Protocol Memory

What is an MCP Server for Memory?

An MCP server for memory is a local service that implements the Model Context Protocol to provide AI coding agents with persistent memory. Instead of losing context between sessions, your agent stores decisions, patterns, and bugs in a searchable memory store.

toon-memory is an MCP server that provides:

• Persistent memory — survives restarts, no cloud required
• 35 MCP tools — full memory management via Model Context Protocol
• Token-efficient format — 22% fewer tokens than JSON
• Graph-aware recall — BM25 + centrality + quality scoring
• 15+ agent support — works with Claude, Cursor, OpenCode, and more

How MCP Memory Server Works

Architecture

Your AI Agent (Claude/Cursor/OpenCode)
        ↓ MCP Protocol (stdio)
  toon-memory MCP Server
        ↓
  .toon-memory/memory/data.toon (local file)

1. Agent sends request — “Remember this decision” or “Search for database patterns”
2. MCP server processes — parses request, manages memory store
3. Memory stored locally — TOON format in .toon-memory/memory/
4. Response returned — formatted for the agent to use

MCP Tools

toon-memory exposes 35 MCP tools and 4 MCP resources:

Tool	Purpose
memory_remember	Save decisions, patterns, bugs, knowledge
memory_recall	Search memory (flat or graph mode)
memory_smart_recall	Unified recall: BM25 + graph + quality
memory_forget	Remove entries
memory_stats	View memory state
memory_summary	Save/retrieve file summaries
memory_archive	Archive old entries
memory_diff	Show changes since date
memory_suggest	Find related entries
memory_sessions	Multi-session coordination
context_brief	One-call context briefing
context_generate	Full project briefing
context_diff	Incremental briefing
context_focus	Hyper-focused briefing
context_health	Memory health audit
context_export	Export as markdown
memory_encrypt	AES-256-GCM encryption
memory_decrypt	Disable encryption
memory_captured	View hook-captured activity
memory_consolidate	Merge duplicates

Why MCP for Memory?

MCP vs Custom Integrations

Feature	MCP Memory	Custom Integration
Standard protocol	Yes (Model Context Protocol)	No — proprietary
Agent support	15+ agents	Usually 1-2 agents
Setup	npx toon-memory	Manual code changes
Updates	npm update	Manual maintenance
Community	Growing ecosystem	Isolated

MCP vs Cloud Memory

Feature	MCP Memory (toon-memory)	Cloud Memory (mem0)
Data location	Local machine	Cloud server
Internet required	No	Yes
Cost	Free	Usage-based
Latency	Instant (local)	Network-dependent
Privacy	Full control	Third-party

Technical Details

TOON Format

Memory is stored in TOON format (Token-Oriented Object Notation):

version: 1
entries[3|]{id|category|key|content|file|tags|date|ttl|accessed|links|quality|confidence}:
  a1b2c3d4|decision|use-zod|Use Zod for validation|src/types.ts|validation;types|2026-07-10||0||0.65|1.0
  e5f6g7h8|pattern|api-routes|REST endpoints use /api/v1/ prefix|src/routes|api;rest;pattern|2026-07-10||0||0.55|1.0
  i9j0k1l2|bug|redis-fix|Redis pool exhaustion fix|src/redis.ts|redis;bug|2026-07-10||0||0.70|1.0

Graph-Aware Recall

Entries are connected via explicit links and implicit references:

memory_remember({
  category: "decision",
  key: "use-redis",
  content: "Use Redis for session storage (see [[redis-config]])",
  links: "redis-config"
})

When you search, recall expands the ego-subgraph to find related entries, even if they don’t contain the exact search term.

Quality Scoring

Every entry gets an automatic quality score (0-1):

Factor	Weight	What it measures
Tags	0.3 max	More specific tags = higher quality
Links	0.2 max	Connected entries = higher quality
Content length	0.3 max	Detailed > vague
Recency	0.1 max	Recent entries score higher
Specificity	0.1 max	Unique words vs repeated words

Get Started

npm install -g toon-memory
npx toon-memory  # Interactive installer

See the Quick Start for detailed setup instructions.