A Node.js server that provides access to the Voiceflow Knowledge Base Query API using the Model Context Protocol (MCP) for seamless AI assistant integration.

1. Clone this repository
2. Install dependencies:

   npm install
   

3. Create a .env file based on the .env.example template:

   cp .env.example .env
   

4. Add your Voiceflow API key and customize settings in the .env file

The server will communicate using the Model Context Protocol over stdio.

This server exposes the following MCP tool:

Query the Voiceflow Knowledge Base using the MCP protocol.

Input Schema:

{
  "question": "What products do you sell?",
  "chunkLimit": 3,
  "synthesis": true,
  "settings": {
    "model": "claude-3-sonnet",
    "temperature": 0.1,
    "system": "You are an AI assistant..."
  }
}

Example requests:

Basic query with default settings:

{
  "question": "What is Voiceflow?"
}

Query with synthesis disabled (returns only knowledge chunks without AI processing):

{
  "question": "How do I create a dialog in Voiceflow?",
  "synthesis": false,
  "chunkLimit": 5
}

Query with custom model settings:

{
  "question": "How do I integrate with Discord?",
  "settings": {
    "model": "gpt-4o",
    "temperature": 0.3,
    "system": "You are a technical support expert. Provide step-by-step instructions."
  }
}

Note: The settings object is only used when synthesis is true. If synthesis is false, the settings are ignored, and only the relevant chunks are returned without a synthesized answer.

To use this server with an MCP-compatible system, configure it as follows:

"vf-kb-server": {
  "command": "node",
  "args": ["index.js"],
  "env": {
    "VOICEFLOW_API_KEY": "${VOICEFLOW_API_KEY}",
    "DEFAULT_SYNTHESIS": "${DEFAULT_SYNTHESIS}",
    "DEFAULT_CHUNK_LIMIT": "${DEFAULT_CHUNK_LIMIT}",
    "DEFAULT_MODEL": "${DEFAULT_MODEL}",
    "DEFAULT_TEMPERATURE": "${DEFAULT_TEMPERATURE}",
    "DEFAULT_SYSTEM_PROMPT": "${DEFAULT_SYSTEM_PROMPT}"
  },
  "disabled": false,
  "autoApprove": []
}

You can omit any of the optional environment variables to use the server's default values.

• VOICEFLOW_API_KEY: Your Voiceflow Knowledge Base API key

• DEFAULT_SYNTHESIS: Whether to synthesize responses by default (true or false, default: false)
• DEFAULT_CHUNK_LIMIT: Default number of chunks to use for response (default: 3)

• DEFAULT_MODEL: Default model to use (default: gpt-4)
• DEFAULT_TEMPERATURE: Default temperature setting (default: 0.1)
• DEFAULT_SYSTEM_PROMPT: Default system prompt for the AI assistant

• Missing API Key: If you get an error about VOICEFLOW_API_KEY not being configured, make sure your .env file contains a valid API key.
• Query Failures: Check your API key permissions and ensure it has access to the Knowledge Base you're trying to query.
• Empty Results: If your queries return no chunks, try increasing the chunk limit or refining your question to be more specific.

The Voiceflow Knowledge Base API allows you to query documents stored in a Voiceflow Knowledge Base. It can return both raw document chunks and AI-synthesized answers based on the retrieved information. For more details, see the Voiceflow Developer Documentation.