Connect Any LLM to Klavis MCP Servers

Connect your favorite LLM to external services like GSuite, CRMs, Slack, and GitHub in minutes with Klavis MCP servers. No complex integrations needed – just plug and play.

Quick Start

Prerequisites

Klavis AI API Key - Get at klavis.ai
API key for your LLM provider (OpenAI, Google AI, Anthropic, etc.)

Basic Setup

All LLM integrations follow the same pattern:

from klavis import Klavis
from klavis.types import McpServerName

# Initialize Klavis client
klavis_client = Klavis(api_key="YOUR_KLAVIS_API_KEY")

# Create an MCP server instance (example: YouTube)
server = klavis_client.mcp_server.create_server_instance(
    server_name=McpServerName.YOUTUBE,
    user_id="user123",
)

print(f"🔗 MCP server ready: {server.server_url}")

Complete Example: OpenAI + YouTube

Here's a working example that lets ChatGPT analyze YouTube videos:

import json
from openai import OpenAI
from klavis import Klavis
from klavis.types import McpServerName, ToolFormat

# Initialize clients
openai_client = OpenAI(api_key="YOUR_OPENAI_API_KEY")
klavis_client = Klavis(api_key="YOUR_KLAVIS_API_KEY")

# Create YouTube MCP server
youtube_server = klavis_client.mcp_server.create_server_instance(
    server_name=McpServerName.YOUTUBE,
    user_id="user123",
)

def analyze_youtube_video(video_url: str):
    """Analyze a YouTube video using OpenAI + Klavis"""
    
    # Get YouTube tools in OpenAI format
    tools = klavis_client.mcp_server.list_tools(
        server_url=youtube_server.server_url,
        format=ToolFormat.OPENAI,
    )
    
    # Ask OpenAI to analyze the video
    response = openai_client.chat.completions.create(
        model="gpt-4o-mini",
        messages=[
            {"role": "system", "content": "You are a helpful assistant. Use available tools to analyze content."},
            {"role": "user", "content": f"Analyze this YouTube video and summarize key points: {video_url}"}
        ],
        tools=tools.tools
    )
    
    # Handle tool calls automatically
    if response.choices[0].message.tool_calls:
        for tool_call in response.choices[0].message.tool_calls:
            function_name = tool_call.function.name
            function_args = json.loads(tool_call.function.arguments)
            
            # Execute tool via Klavis
            result = klavis_client.mcp_server.call_tools(
                server_url=youtube_server.server_url,
                tool_name=function_name,
                tool_args=function_args,
            )
            
            # Return the analysis
            return f"Video Analysis: {result.result.content[0]['text']}"
    
    return response.choices[0].message.content

# Test it
video_url = "https://www.youtube.com/watch?v=dQw4w9WgXcQ"
analysis = analyze_youtube_video(video_url)
print(analysis)

Works with Any LLM

The same pattern works with any LLM provider. Just change the tool format:

Google Gemini

# Get tools in Gemini format
tools = klavis_client.mcp_server.list_tools(
    server_url=server.server_url,
    format=ToolFormat.GEMINI,
)

Anthropic Claude

# Get tools in Claude format
tools = klavis_client.mcp_server.list_tools(
    server_url=server.server_url,
    format=ToolFormat.ANTHROPIC,
)

Custom LLMs

# Get tools in generic format for custom implementations
tools = klavis_client.mcp_server.list_tools(
    server_url=server.server_url,
    format=ToolFormat.GENERIC,
)

OAuth Services (Gmail, Slack, etc.)

For services requiring OAuth authentication:

import webbrowser

# Create Gmail MCP server
gmail_server = klavis_client.mcp_server.create_server_instance(
    server_name=McpServerName.GMAIL,
    user_id="user123",
)

# Open OAuth URL for user authorization
webbrowser.open(gmail_server.oauth_url)
print(f"🔐 Complete authorization: {gmail_server.oauth_url}")

# After OAuth completion, use with any LLM