Connect Your Agent

Add persistent memory to your AI agent in under 2 minutes. MemLayer works with any MCP-compatible client via stdio transport.

1. Start memlayer

Make sure memlayer is running:

bb local-server    # local mode (recommended, port 8090)
# or: bb server    # development mode (port 8080)

2. Configure your agent

Pick your client and add memlayer as an MCP server:

Claude Code

One command (recommended):

claude mcp add memlayer -- bb mcp

Or create .mcp.json in your project root:

{
  "mcpServers": {
    "memlayer": {
      "command": "bb",
      "args": ["mcp"],
      "cwd": "/path/to/memlayer"
    }
  }
}

Reload your Claude Code session to pick up the config.

Claude Desktop

Edit your config file:

OS	Path
macOS	~/Library/Application Support/Claude/claude_desktop_config.json
Windows	%APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "memlayer": {
      "command": "bb",
      "args": ["mcp"],
      "cwd": "/path/to/memlayer"
    }
  }
}

Restart Claude Desktop after saving.

Cursor

Create .cursor/mcp.json in your project root:

{
  "mcpServers": {
    "memlayer": {
      "command": "bb",
      "args": ["mcp"],
      "cwd": "/path/to/memlayer"
    }
  }
}

Or go to Cursor Settings > MCP > Add new MCP server and configure the command.

Windsurf

Create .windsurf/mcp.json in your project root:

{
  "mcpServers": {
    "memlayer": {
      "command": "bb",
      "args": ["mcp"],
      "cwd": "/path/to/memlayer"
    }
  }
}

Or configure via Windsurf Settings > MCP.

VS Code Copilot

Create .vscode/mcp.json in your project root:

{
  "servers": {
    "memlayer": {
      "type": "stdio",
      "command": "bb",
      "args": ["mcp"],
      "cwd": "/path/to/memlayer"
    }
  }
}

Or go to Settings > GitHub Copilot > MCP and add the server.

Cline

Open Cline’s MCP settings panel in VS Code and add a new server with name memlayer, transport stdio, command bb, and args ["mcp"]. Set the working directory to your memlayer installation path.

3. Verify

Ask your agent:

“Remember that I prefer dark mode in all my editors”

If memlayer is connected, the agent will call the memlayer_retain tool and confirm the memory was stored.

Then test recall:

“What are my editor preferences?”

The agent calls memlayer_recall and returns your stored memory.

What tools does my agent get?

Once connected, your agent has access to five tools:

Tool	What it does
memlayer_retain	Store memories from conversation content
memlayer_recall	Semantic search across stored memories
memlayer_reflect	Consolidate scattered memories into concepts
memlayer_forget	Remove a memory from active queries
memlayer_batch_retain	Store multiple memories in a single call

See MCP Tools Reference for full parameter details.

How the agent knows to use memlayer

When your agent connects via MCP, two things happen:

1. Tool discovery: The agent receives tool definitions (names, parameters, descriptions) through the MCP tools/list method. This tells the agent what tools are available.

2. Behavioral instructions via SKILL.md: memlayer serves a SKILL.md file that your agent loads as a system prompt. This is the key part — it teaches the agent when and how to use memory effectively. The instructions include:

   • When to recall: Always check memory before answering questions about the user’s preferences, past decisions, or project context. Do this silently — don’t announce it.
   • When to retain: Watch for high-signal moments like explicit preferences (“I prefer Tailwind”), corrections, decisions, and project context. Retain immediately and unobtrusively.
   • How to write good memories: Third person, specific, self-contained, one fact per memory.
   • Budget awareness: Be conservative — fewer high-quality memories beat many low-value ones.
   • End of conversation: Review for any unretained facts worth saving and propose them to the user.

   Without SKILL.md, the agent would have the tools but wouldn’t know the best practices for using them. The skill file is what makes memory feel natural rather than mechanical.

   You can read the full instructions at resources/public/SKILL.md in the repository.

Other MCP clients

Any client that supports stdio transport can connect to memlayer:

Setting	Value
Command	bb
Args	["mcp"]
Working directory	Path to your memlayer installation
Transport	stdio

Next steps

• Quickstart: MCP — guided walkthrough of all four operations
• Quickstart: API — use the REST API directly with curl
• Concepts — understand how memlayer organizes knowledge