> ## Documentation Index
> Fetch the complete documentation index at: https://docs.spikeapi.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Desktop Chat Clients

> Local testing setup for exploring Spike MCP capabilities through desktop chat clients.

<Warning>
  This guide is for **local testing and experimentation only**. For production integrations, see the [Implementation Guide](/mcp-docs/implementation) for complete examples with OpenAI and Anthropic.
</Warning>

Desktop chat clients like Claude Desktop provide an easy way to test MCP capabilities and experiment with health data queries locally.

## Claude Desktop Setup

Download Claude Desktop from [https://claude.ai/download](https://claude.ai/download).

### Step 1: Install MCP Proxy

Install the open-source MCP proxy tool:

```bash theme={null}
brew install mcp-proxy
```

### Step 2: Configure Claude Desktop

1. Open Claude Desktop and navigate to **Settings** → **Developer**
2. Click the button to edit the configuration file

<Frame>
  <img src="https://mintcdn.com/trial-7e3c8695/tLxMhymXagiR2q29/images/claude_desktop_config.png?fit=max&auto=format&n=tLxMhymXagiR2q29&q=85&s=7cb8bd065efb620819fb74de8cfd1732" alt="Claude config menus" width="2352" height="1480" data-path="images/claude_desktop_config.png" />
</Frame>

3. Add the following configuration to the file (replace `<SPIKE_ACCESS_TOKEN>` with your actual token):

```json theme={null}
{
  "mcpServers": {
    "mcp-proxy": {
      "command": "mcp-proxy",
      "args": [
        "-H",
        "Authorization",
        "Bearer <SPIKE_ACCESS_TOKEN>",
        "--transport",
        "streamablehttp",
        "https://app-api.spikeapi.com/v3/mcp"
      ]
    }
  }
}
```

4. Save the file and restart Claude Desktop

### Step 3: Test the Connection

Try a simple query to verify everything is working:

```
How well did I sleep last night?
```

You should see the `query_sleep` tool being used and may be prompted for approval:

<Frame>
  <img src="https://mintcdn.com/trial-7e3c8695/tLxMhymXagiR2q29/images/claude_desktop_query.png?fit=max&auto=format&n=tLxMhymXagiR2q29&q=85&s=ca86de390171bdb48cb4160eb114f38e" alt="Claude config menus" width="2374" height="1164" data-path="images/claude_desktop_query.png" />
</Frame>

Allow the connection to get personalized results based on your JWT token:

<Frame>
  <img src="https://mintcdn.com/trial-7e3c8695/tLxMhymXagiR2q29/images/claude_desktop_result.png?fit=max&auto=format&n=tLxMhymXagiR2q29&q=85&s=4f3bcc54b3d3a6a0a4ff75dceaba8517" alt="Claude config menus" width="2374" height="1234" data-path="images/claude_desktop_result.png" />
</Frame>

## Next Steps

For production integrations and complete implementation examples, see the [Implementation Guide](/mcp-docs/implementation) which includes:

* Complete OpenAI integration examples (Python, Go, Node.js, cURL)
* Anthropic/Claude integration examples
* Authentication setup
* Error handling
* Available MCP tools and endpoints