Claude Desktop Integration

Give Claude persistent memory with AgentRecall. Works with the Claude Desktop app on macOS and Windows.

Prerequisites

• Claude Desktop app installed (download)
• An AgentRecall API key (get one free)
• Python 3.10+ installed

Quick Setup

1. Create an agent and API key

Go to the dashboard and:

1. Create a new agent (e.g., "claude-desktop")
2. Create an API key for that agent
3. Copy the key (starts with ark_)

2. Install the MCP server

pip install agentrecall-mcp

3. Configure Claude Desktop

Open your Claude Desktop config file:

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json

Add the AgentRecall server:

{
  "mcpServers": {
    "agentrecall": {
      "command": "agentrecall-mcp",
      "env": {
        "AGENTRECALL_API_KEY": "ark_YOUR_KEY_HERE"
      }
    }
  }
}

Note: You don't need to specify AGENTRECALL_AGENT_ID — the API key is already bound to your agent.

4. Restart Claude Desktop

Quit and reopen the app. You should see a 🔨 icon in the input area showing AgentRecall tools are connected.

What You Get

Once connected, Claude can:

• Remember things — Store information across conversations
• Search memories — Find relevant context from past chats
• Build a knowledge graph — Connect related concepts and entities
• Traverse relationships — Explore how memories connect

Try It

After setup, try asking Claude:

• "Remember that my favorite programming language is Rust"
• "What do you remember about my preferences?"
• "Search for any memories about programming"

Troubleshooting

No tools appearing

• Check that agentrecall-mcp is in your PATH: which agentrecall-mcp
• Verify the JSON config is valid (no trailing commas)
• Check Claude Desktop logs: ~/Library/Logs/Claude/ (macOS)

Connection errors

• Make sure your API key starts with ark_
• Test the key: curl -H "Authorization: Bearer ark_YOUR_KEY" https://api.agentrecall.cloud/v1/health

Multiple Agents

Each API key is bound to one agent. To use multiple agents (e.g., personal vs work), create separate agents in the dashboard and configure separate MCP instances with different keys.