Bland.ai MCP Server

A Model Context Protocol (MCP) server that provides read-only access to Bland.ai API call data.

Features

This MCP server provides the following tools:

1. get-call-details - Retrieve complete call information including transcript from Bland.ai
2. list-calls - List Bland.ai calls with optional filters for pathway, date range, and limit

Quick Start (No Installation Required!)

The easiest way to use this MCP server is directly with npx - no installation needed!

Setup for Claude Desktop

1. Get your Bland.ai API Key

Get your API key from your Bland.ai dashboard.

2. Configure Claude Desktop

Add this to your Claude Desktop configuration file:

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

{
  "mcpServers": {
    "bland-mcp": {
      "command": "npx",
      "args": ["@tgerighty/bland-mcp"],
      "env": {
        "BLAND_AI_API_KEY": "your_actual_bland_api_key_here"
      }
    }
  }
}

That's it! No installation needed. The server will be downloaded and run automatically when Claude starts.

3. Restart Claude Desktop

After saving the configuration, restart Claude Desktop completely.

4. Verify It's Working

In Claude, you can now use commands like:

• "List my recent Bland.ai calls"
• "Get details for call [call-id]"
• "Show calls from the last 24 hours"

Alternative Installation Methods

Install from GitHub (for development):

{
  "mcpServers": {
    "bland-mcp": {
      "command": "node",
      "args": ["/path/to/bland-mcp/dist/index.js"],
      "env": {
        "BLAND_AI_API_KEY": "your_actual_bland_api_key_here"
      }
    }
  }
}

3. Restart Claude Desktop

After saving the configuration, restart Claude Desktop completely.

4. Verify Installation

In Claude, you can now use commands like:

• "List my recent Bland.ai calls"
• "Get details for call [call-id]"
• "Show calls from the last 24 hours"

Usage

Testing the Server

To test the server in standalone mode:

node mcp-server.js

This will show the available tools without starting the MCP protocol.

Running as MCP Server

To run as an MCP server:

node mcp-server.js --mcp

Claude Desktop Integration

To use with Claude Desktop, you need to add this server to your MCP configuration. Edit the Claude Desktop configuration file:

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

Add the following to the mcpServers section:

Option 1: Using npx (Recommended)

{
  "mcpServers": {
    "bland-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "/Users/tgerighty/Library/CloudStorage/OneDrive-SignantHealth/Documents/Coding/bland-mcp"
      ],
      "env": {
        "BLAND_AI_API_KEY": "your_actual_bland_api_key_here"
      }
    }
  }
}

Option 2: Using full path with node

{
  "mcpServers": {
    "bland-mcp": {
      "command": "node",
      "args": [
        "/Users/tgerighty/Library/CloudStorage/OneDrive-SignantHealth/Documents/Coding/bland-mcp/dist/index.js"
      ],
      "env": {
        "BLAND_AI_API_KEY": "your_actual_bland_api_key_here"
      }
    }
  }
}

Replace your_actual_bland_api_key_here with your actual Bland.ai API key.

For detailed integration instructions, see INTEGRATION-GUIDE.md.

API Endpoints

The tools access the following Bland.ai API endpoints:

• get-call-details:

  • /v1/calls/{call_id} - Complete call information including transcript

• list-calls:

  • /v1/calls - List all calls
  • Optional query parameters: pathway_id, from_date, to_date, limit

Environment Variables

• BLAND_AI_API_KEY - Your Bland.ai API key (required)
• PORT - Port for the web server (default: 3000)

Error Handling

All tools will return error information if API calls fail, including the error message from the Bland.ai API.

Troubleshooting

list-calls API returns empty results from Node.js

When attempting to use the list-calls tool from a Node.js environment (e.g., via axios or native fetch), the API might return an empty calls array, even if total_count indicates that calls exist. This issue has been observed despite matching request headers (including User-Agent and Accept) and attempting to force HTTP/2.

Workaround:

A direct curl command from the same environment successfully retrieves the call data. If you encounter this issue, consider executing curl commands directly from your Node.js application using child_process.exec for list-calls functionality.

Example of a working curl command:

curl --request GET \
  --url "https://api.bland.ai/v1/calls?limit=5" \
  --header "authorization: YOUR_BLAND_API_KEY"

This suggests a subtle difference in how the Bland.ai API responds to requests originating from Node.js's network stack compared to curl, or an undocumented API behavior. Further investigation into the underlying network differences or direct consultation with Bland.ai API support may be required to fully resolve this.