Claude Desktop Integration

Give Claude persistent, encrypted memory in 3 steps. Every conversation builds on previous context — preferences, decisions, and project history are remembered across sessions.

Overview

Claude Desktop supports the Model Context Protocol (MCP), allowing Claude to use external tools. By connecting Synapse Layer, Claude gains 13 memory tools including store_memory, recall_memory, and search.

Two connection options: Remote (Streamable HTTP, no install) or Local (uvx subprocess, offline capable).

Option A: Remote (Recommended)

Open Claude Desktop settings and add this to your MCP configuration:

json
{
  "mcpServers": {
    "synapse-layer": {
      "url": "https://forge.synapselayer.org/api/mcp",
      "transport": "streamable-http",
      "headers": {
        "x-connect-token": "sk_connect_your_token"
      }
    }
  }
}

Note

Replace sk_connect_your_token with your actual Connect Token from forge.synapselayer.org → Dashboard → Connect.

Option B: Local (uvx)

Requires Python 3.10+ and uv installed locally:

json
{
  "mcpServers": {
    "synapse-layer": {
      "command": "uvx",
      "args": ["synapse-layer"],
      "env": {
        "SYNAPSE_API_KEY": "sk_live_your_api_key",
        "SYNAPSE_AGENT_ID": "claude-desktop"
      }
    }
  }
}

Verify It Works

After restarting Claude Desktop, ask Claude:

"Run health_check to verify Synapse Layer is connected."

Then test memory storage:

"Remember that I prefer concise, technical responses with code examples."

In a new conversation, ask Claude to recall your preferences — it will retrieve the stored memory.

Usage Tips

  • Explicit save — Tell Claude "save this to memory" for important decisions or preferences
  • Cross-session recall — Start new conversations with "recall my project context" to load previous state
  • Agent isolation — Use different SYNAPSE_AGENT_ID values for different projects
  • PII safety — Sensitive data (emails, phone numbers, IPs) is automatically detected and redacted before storage