AgenticMemory MCP Server

Universal LLM access to persistent graph memory via the Model Context Protocol.

AgenticMemory MCP Server bridges any MCP-compatible LLM client (Claude, GPT, Gemini, Ollama, etc.) to the AgenticMemory persistent binary graph memory system. One server, universal access.

Architecture

┌─────────────────────────────────────────────────────────┐
│                     LLM CLIENT                           │
│       (Claude, GPT, Gemini, Ollama, etc.)               │
└─────────────────────┬───────────────────────────────────┘
                      │ MCP Protocol (JSON-RPC 2.0)
                      │ (stdio / SSE)
┌─────────────────────▼───────────────────────────────────┐
│                AGENTIC-MEMORY-MCP                        │
│  ┌──────────┐  ┌──────────┐  ┌──────────┐              │
│  │  TOOLS   │  │RESOURCES │  │ PROMPTS  │              │
│  │ 12 tools │  │ 6 URIs   │  │ 4 tmpls  │              │
│  └────┬─────┘  └────┬─────┘  └────┬─────┘              │
│       └──────────────┼─────────────┘                    │
│              SESSION MANAGER                             │
│              AGENTIC-MEMORY CORE                         │
└─────────────────────────────────────────────────────────┘
                      │
                      ▼
               brain.amem (binary graph file)

Quick Start

Install from crates.io

cargo install agentic-memory-mcp

Or build from source

cargo build --release

Run (stdio transport — default)

agentic-memory-mcp serve

Zero-config: defaults to ~/.brain.amem. Override with agentic-memory-mcp --memory /path/to/brain.amem serve.

Memory Modes

Control how aggressively Claude saves memories:

agentic-memory-mcp serve --mode minimal   # Only save when user says "remember"
agentic-memory-mcp serve --mode smart     # Auto-save facts + decisions (default)
agentic-memory-mcp serve --mode full      # Save everything, episode summaries

Mode	Saves Per Conversation	File Growth	Best For
minimal	1-3	~1 KB/day	Privacy-conscious users
smart	5-15	~5-20 KB/day	Most users (default)
full	15-30+	~20-50 KB/day	Power users, autonomous agents

Configure via the args array in your MCP config:

{ "command": "agentic-memory-mcp", "args": ["serve", "--mode", "full"] }

Run (SSE transport)

cargo build --release --features sse
agentic-memory-mcp serve-http --addr 127.0.0.1:3000

Configuration with MCP Clients

Claude Desktop

Add to ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "agentic-memory": {
      "command": "agentic-memory-mcp",
      "args": ["serve"]
    }
  }
}

Claude Code

Add to ~/.claude/mcp.json:

{
  "mcpServers": {
    "agentic-memory": {
      "command": "agentic-memory-mcp",
      "args": ["serve"]
    }
  }
}

VS Code / Cursor

Add to .vscode/settings.json:

{
  "mcp.servers": {
    "agentic-memory": {
      "command": "agentic-memory-mcp",
      "args": ["serve"]
    }
  }
}

Windsurf

Add to ~/.codeium/windsurf/mcp_config.json:

{
  "mcpServers": {
    "agentic-memory": {
      "command": "agentic-memory-mcp",
      "args": ["serve"]
    }
  }
}

Zero-config: defaults to ~/.brain.amem. Override with "args": ["--memory", "/path/to/brain.amem", "serve"].

Tools (12)

Tool	Purpose
`memory_add`	Add cognitive event to graph
`memory_query`	Pattern query for matching nodes
`memory_traverse`	Walk graph following edges
`memory_correct`	Record correction to past belief
`memory_resolve`	Follow supersedes chain
`memory_context`	Get subgraph around node
`memory_similar`	Similarity search
`memory_causal`	Impact analysis
`memory_temporal`	Compare across time
`memory_stats`	Graph statistics
`session_start`	Begin new session
`session_end`	End session, create episode

Resources (6)

URI Pattern	Returns
`amem://node/{id}`	Single node with edges
`amem://session/{id}`	All nodes from session
`amem://types/{type}`	All nodes of type
`amem://graph/stats`	Graph statistics
`amem://graph/recent`	Recent nodes
`amem://graph/important`	High decay score nodes

Prompts (4)

Prompt	Purpose	Required Args
`remember`	Guide for storing new info	`information`
`reflect`	Guide for understanding past decisions	`topic`
`correct`	Guide for updating beliefs	`old_belief`, `new_information`
`summarize`	Guide for session summary	(none)

Event Types

Fact — Declarative knowledge
Decision — Choices made with reasoning
Inference — Derived conclusions
Correction — Updates to prior beliefs
Skill — Procedural knowledge
Episode — Session summaries

Edge Types

CausedBy — Causal relationship
Supports — Supporting evidence
Contradicts — Conflicting information
Supersedes — Correction chain
RelatedTo — General association
PartOf — Hierarchical containment
TemporalNext — Temporal ordering

CLI Commands

# Start server (stdio) — defaults to ~/.brain.amem
agentic-memory-mcp serve

# Start server with custom memory file
agentic-memory-mcp --memory /path/to/brain.amem serve

# Start server (SSE, requires --features sse)
agentic-memory-mcp serve-http --addr 127.0.0.1:3000

# Validate a memory file
agentic-memory-mcp --memory ~/.brain.amem validate

# Print server info as JSON
agentic-memory-mcp info

Development

This crate is part of the AgenticMemory Cargo workspace.

# Run MCP server tests (from workspace root)
cargo test -p agentic-memory-mcp

# Run bridge integration tests
cargo test -p agentic-memory-bridge-tests

# Run all workspace tests
cargo test --workspace

# Clippy + format
cargo clippy --workspace
cargo fmt --all

# Build release
cargo build --release

Protocol

This server implements MCP (Model Context Protocol) spec version 2024-11-05 over JSON-RPC 2.0. Supported transports:

stdio — Newline-delimited JSON over stdin/stdout (default)
SSE — Server-Sent Events over HTTP (feature flag sse)

License

MIT

agentic-memory-mcp 0.4.1