Your AI agent forgets everything.
This server fixes that.

What changes once it is connected

The agent stops living inside a single chat. Decisions, lessons and preferences survive /clear, survive the week-end, and — most importantly — survive the jump to your next project.

Quick start

dotnet tool install --global Snail.MCP.Memory --prerelease

claude mcp add mem0 --scope user -- snail-mcp-memory

Connecting to a client

The server speaks stdio: every client config boils down to "run snail-mcp-memory" (or dnx Snail.MCP.Memory --yes --prerelease to skip installation). The examples register it as mem0 — the slash commands in the repo pre-approve tools under that name. Zero configuration required — settings only change the defaults.

Claude Code — .mcp.json

{
  "mcpServers": {
    "mem0": {
      "command": "snail-mcp-memory"
    }
  }
}

Running through dnx instead: "command": "dnx", "args": ["Snail.MCP.Memory", "--yes", "--prerelease"].

Claude Desktop

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json · Windows: %APPDATA%\Claude\claude_desktop_config.json — same JSON as above. If the command is not found, use the full path /Users/<you>/.dotnet/tools/snail-mcp-memory.

VS Code — .vscode/mcp.json

{
  "servers": {
    "mem0": {
      "type": "stdio",
      "command": "snail-mcp-memory"
    }
  }
}

Configuration

Every setting is an optional environment variable. This example lists all of them with their defaults — copy it and keep only the lines you change:

{
  "mcpServers": {
    "mem0": {
      "command": "snail-mcp-memory",
      "env": {
        "SNAIL_MCP_MEMORY_PROVIDER": "Sqlite",
        "SNAIL_MCP_MEMORY_DIRECTORY": "~/.snail-mcp-memory",
        "SNAIL_MCP_MEMORY_CONNECTION_STRING": "",
        "SNAIL_MCP_MEMORY_WORKSPACE": "",
        "SNAIL_MCP_MEMORY_EMBEDDING_ENABLED": "true",
        "SNAIL_MCP_MEMORY_EMBEDDING_MODEL_PATH": "~/.snail-mcp-memory/models/model.onnx",
        "SNAIL_MCP_MEMORY_EMBEDDING_TOKENIZER_PATH": "~/.snail-mcp-memory/models/sentencepiece.bpe.model",
        "SNAIL_MCP_MEMORY_EMBEDDING_MODEL_URL": "https://huggingface.co/Xenova/multilingual-e5-small/resolve/main/onnx/model_quantized.onnx",
        "SNAIL_MCP_MEMORY_EMBEDDING_TOKENIZER_URL": "https://huggingface.co/intfloat/multilingual-e5-small/resolve/main/sentencepiece.bpe.model",
        "SNAIL_MCP_MEMORY_EMBEDDING_DIMENSIONS": "384",
        "SNAIL_MCP_MEMORY_DUPLICATE_THRESHOLD": "0.95",
        "SNAIL_MCP_MEMORY_SIMILARITY_FLOOR": "0.75",
        "SNAIL_MCP_MEMORY_IDLE_TIMEOUT_MINUTES": "60"
      }
    }
  }
}

The same keys — bare names, without the prefix — can live in a JSON file instead: ~/.snail-mcp-memory.json (also read: the path in SNAIL_MCP_MEMORY_CONFIG, ./.snail-mcp-memory.json, ./snail-mcp-memory.json, ~/.config/snail-mcp-memory/config.json). Environment variables win over the file.

PostgreSQL

For a shared or backed-up memory, point the server at PostgreSQL with the pgvector extension (the pgvector/pgvector Docker images ship it). Nothing else changes — same tools, same behavior:

"SNAIL_MCP_MEMORY_PROVIDER": "Postgres",
"SNAIL_MCP_MEMORY_CONNECTION_STRING": "Host=db.example.com;Database=memory;Username=snail;Password=…"

What the agent can do with it

You normally never call these yourself — the agent does, guided by instructions the server ships inside the protocol itself. Knowing them helps you phrase requests.

Every record carries a kind; the vocabulary is open, the core carries semantics:

Integration: copy two things from the repo

The protocol carries the tool descriptions — enough for basic use. The full integration is two files you copy yourself from the git repository into your agent's setup; nothing is installed automatically:

1. Slash commands. .claude/commands/ holds mem0-init, mem0-recall, mem0-remember, mem0-link, mem0-graph, mem0-list — manual one-shot shortcuts with the mem0 tools pre-approved, so they run without permission prompts:

cp .claude/commands/mem0-*.md ~/.claude/commands/

Registered the server under another name? Adjust the mcp__mem0__ prefix inside the files.

2. Memory rules. server-instructions.md at the repo root is the full instruction set — the kind vocabulary, when to save unprompted, when to search before proposing, how to self-heal wrong records. Paste its content into your agent's instruction file — ~/.claude/CLAUDE.md for Claude Code, or the system prompt of a custom agent. The agent works without it, but with it the memory maintains itself.

Sharing one database

Good to know