Getting started with Knit-MCP servers for efficient workflow management.

What is Klavis AI?

Klavis AI is open source MCP integrations for AI Applications. Our API provides hosted, high quality, secure MCP servers, eliminating auth management and client-side code.\n\n## ✨ Key Features

• 🚀 Instant Integration: Get started in minutes with our Python and TypeScript SDKs, or simply REST API
• 🔐 Built-in Authentication: Secure OAuth flows and API key management
• ⚡ Production-Ready: Hosted infrastructure that scales to millions of users
• 🛠️ 100+ Tools: Access to CRM, GSuite, Github, Slack, databases, and many more
• 🌐 Multi-Platform: Works with any LLM provider (OpenAI, Anthropic, Gemini, etc.) and any AI agent framework (LangChain, Llamaindex, CrewAI, AutoGen, etc.)
• 🔧 Self-Hostable: Open-source MCP servers you can run yourself

🚀 Quick Start# Installation

Python

pip install klavis

TypeScript/JavaScript

npm install klavis

Get Your API Key

Sign up at klavis.ai and create your API key.

With MCP Client

If you already have an MCP client implementation in your codebase:

Python Example

from klavis import Klavis
from klavis.types import McpServerName, ConnectionType

klavis_client = Klavis(api_key="your-klavis-key")

# Create a YouTube MCP server instance
youtube_server = klavis_client.mcp_server.create_server_instance(
    server_name=McpServerName.YOUTUBE,
    user_id="user123", # Change to user id in your platform
    platform_name="MyApp" # change to your platform
)

print(f"Server created: {youtube_server.server_url}")

TypeScript Example

import { KlavisClient, Klavis } from 'klavis';

const klavisClient = new KlavisClient({ apiKey: 'your-klavis-key' });

// Create Gmail MCP server with OAuth
const gmailServer = await klavisClient.mcpServer.createServerInstance({
    serverName: Klavis.McpServerName.Gmail,
    userId: "user123",
    platformName: "MyApp"
});

// Gmail needs OAuth flow
await window.open(gmailServer.oauthUrl);

Without MCP Client (Function Calling)

Integrate directly with your LLM provider or AI agent framework using function calling:

Python + OpenAI Example

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

OPENAI_MODEL = "gpt-4o-mini"

openai_client = OpenAI(api_key="YOUR_OPENAI_API_KEY")
klavis_client = Klavis(api_key="YOUR_KLAVIS_API_KEY")

# Create server instance
youtube_server = klavis_client.mcp_server.create_server_instance(
    server_name=McpServerName.YOUTUBE,
    user_id="user123",
    platform_name="MyApp"
)

# Get available tools in OpenAI format
tools = klavis_client.mcp_server.list_tools(
    server_url=youtube_server.server_url,
    format=ToolFormat.OPENAI,
)

# Initial conversation
messages = [{"role": "user", "content": "Summarize this video: https://youtube.com/watch?v=..."}]

# First OpenAI call with function calling
response = openai_client.chat.completions.create(
    model=OPENAI_MODEL,
    messages=messages,
    tools=tools.tools
)

messages.append(response.choices[0].message)

# Handle tool calls
if response.choices[0].message.tool_calls:
    for tool_call in response.choices[0].message.tool_calls:
        result = klavis_client.mcp_server.call_tools(
            server_url=youtube_server.server_url,
            tool_name=tool_call.function.name,
            tool_args=json.loads(tool_call.function.arguments),
        )
        
        # Add tool result to conversation
        messages.append({
            "role": "tool",
            "tool_call_id": tool_call.id,
            "content": str(result)
        })

# Second OpenAI call to process tool results and generate final response
final_response = openai_client.chat.completions.create(
    model=OPENAI_MODEL,
    messages=messages
)

print(final_response.choices[0].message.content)

TypeScript + OpenAI Example

import OpenAI from 'openai';
import { KlavisClient, Klavis } from 'klavis';

// Constants
const OPENAI_MODEL = "gpt-4o-mini";

const EMAIL_RECIPIENT = "[email protected]";
const EMAIL_SUBJECT = "Hello from Klavis";
const EMAIL_BODY = "This email was sent using Klavis MCP Server!";

const openaiClient = new OpenAI({ apiKey: 'your-openai-key' });
const klavisClient = new KlavisClient({ apiKey: 'your-klavis-key' });

// Create server and get tools
const gmailServer = await klavisClient.mcpServer.createServerInstance({
    serverName: Klavis.McpServerName.Gmail,
    userId: "user123",
    platformName: "MyApp"
});

// Handle OAuth authentication for Gmail
if (gmailServer.oauthUrl) {
    console.log("Please complete OAuth authorization:", gmailServer.oauthUrl);
    await window.open(gmailServer.oauthUrl);
}

const tools = await klavisClient.mcpServer.listTools({
    serverUrl: gmailServer.serverUrl,
    format: Klavis.ToolFormat.Openai
});

// Initial conversation
const messages = [{ 
    role: "user", 
    content: `Please send an email to ${EMAIL_RECIPIENT} with subject "${EMAIL_SUBJECT}" and body "${EMAIL_BODY}"` 
}];

// First OpenAI call with function calling
const response = await openaiClient.chat.completions.create({
    model: OPENAI_MODEL,
    messages: messages,
    tools: tools.tools
});

messages.push(response.choices[0].message);

// Handle tool calls
if (response.choices[0].message.tool_calls) {
    for (const toolCall of response.choices[0].message.tool_calls) {
        const result = await klavisClient.mcpServer.callTools({
            serverUrl: gmailServer.serverUrl,
            toolName: toolCall.function.name,
            toolArgs: JSON.parse(toolCall.function.arguments)
        });
        
        // Add tool result to conversation
        messages.push({
            role: "tool",
            tool_call_id: toolCall.id,
            content: JSON.stringify(result)
        });
    }
}

// Second OpenAI call to process tool results and generate final response
const finalResponse = await openaiClient.chat.completions.create({
    model: OPENAI_MODEL,
    messages: messages
});

console.log(finalResponse.choices[0].message.content);

📚 AI Platform Integration Tutorials

• AI Platform Integrations Overview - Learn how to integrate with leading AI platforms
• Together AI Integration - Build AI agents with Together AI's high-performance LLMs
• OpenAI Integration - Create fast and efficient AI agents with OpenAI and Klavis MCP Servers
• And More!

🛠️ Available MCP Servers

View All Servers →

🗺️ Roadmap

• more high quality MCP Servers (Figma, Canva, Perplexityai, Microsoft teams, Google maps, Bitbucket, Cloudflare, Zoho, Tavily, Posthog, Mem0, Brave Search, Apollo, Exa, Fireflies, Eleven Labs, Hacker News, MicroSoft Calendar)
• More AI platform integrations (examples & docs )
• Event-driven / Webhook
• Unit Tests, integration test
• /docs improvement

🔧 Authentication & Multi-Tool Workflows# Authentication

Many MCP servers require authentication. Klavis handles this seamlessly:

server = klavis_client.mcp_server.create_server_instance(
    server_name=McpServerName.GMAIL,
    user_id="user123",
    platform_name="MyApp"
)\n\n# Option 1 - OAuth URL is provided in server.oauth_url, redirect user to OAuth URL for authentication
import webbrowser
webbrowser.open(server.oauth_url)

# Option 2 - or for API key services
klavis_client.mcp_server.set_auth_token(
    instance_id=server.instance_id,
    auth_token="your-service-api-key"
)

🏠 Self-Hosting

Want to run MCP servers yourself? All our servers are open-source:

git clone https://github.com/klavis-ai/klavis.git
cd klavis

# Run a specific MCP server
cd mcp_servers/github
docker build -t klavis-github .
docker run -p 8000:8000 klavis-github

checkout each readme for more details

📖 Documentation

• API Documentation - Complete API reference
• SDK Documentation - Python & TypeScript guides
• MCP Protocol Guide - Understanding MCP
• Authentication Guide - OAuth and API keys

🤝 Contributing

We welcome contributions! Here's how to get started:

1. Report Issues: Found a bug? Open an issue
2. Request Features: Have an idea? Start a discussion
3. Contribute Code: Check our Contributing Guidelines
4. Build MCP Servers: Want to add new integrations? See our MCP Server Guide
5. Join Community: Connect with us on Discord

📜 License

This project is licensed under the MIT License - see the LICENSE file for details.