Klavis AI – open-source MCP servers and SDKs that let you securely call 100+ SaaS/DB/infra tools from LLMs or agent frameworks with zero auth plumbing.
https://github.com/Klavis-AI/klavisYou know the drill. You want to build an AI agent that can send emails, create GitHub issues, update spreadsheets, or manage your CRM. Then you spend the next three weeks fighting OAuth flows, parsing API docs, and handling rate limits instead of actually building your agent.
Klavis AI cuts through this integration mess with pre-built MCP servers for 100+ popular services. No more authentication headaches, no more API wrapper code, no more "it works in development but breaks in production" moments.
Building agents that can actually do things requires connecting to external services. That means:
You end up spending 80% of your time on plumbing and 20% on the actual AI logic. Klavis flips this ratio.
Instead of building integrations from scratch, you get production-ready MCP servers with everything handled:
from klavis import Klavis
from openai import OpenAI
# Create a Gmail server instance - OAuth handled automatically
klavis = Klavis(api_key="your-key")
gmail_server = klavis.mcp_server.create_server_instance(
server_name="gmail",
user_id="user123",
platform_name="MyApp"
)
# Get tools for your LLM - properly formatted, ready to use
tools = klavis.mcp_server.list_tools(
server_url=gmail_server.server_url,
format="openai"
)
# Your agent can now send emails, search messages, manage labels
openai_client = OpenAI()
response = openai_client.chat.completions.create(
model="gpt-4o",
messages=[{"role": "user", "content": "Send a follow-up email to [email protected]"}],
tools=tools.tools
)
The OAuth flow is handled through gmail_server.oauth_url. The tool schemas are automatically generated. Error handling and retries are built-in. You focus on your agent logic, not API integration details.
Gmail + Calendar + Slack workflow in 15 minutes:
# Create multiple servers
gmail = klavis.mcp_server.create_server_instance(server_name="gmail", ...)
calendar = klavis.mcp_server.create_server_instance(server_name="google_calendar", ...)
slack = klavis.mcp_server.create_server_instance(server_name="slack", ...)
# Combine all tools
all_tools = []
all_tools.extend(klavis.mcp_server.list_tools(gmail.server_url).tools)
all_tools.extend(klavis.mcp_server.list_tools(calendar.server_url).tools)
all_tools.extend(klavis.mcp_server.list_tools(slack.server_url).tools)
# Agent can now coordinate across all three services
messages = [{"role": "user", "content": "Schedule a meeting with the team and send a Slack reminder"}]
Compare this to building the same functionality from scratch:
Klavis covers the services developers actually use in production:
Developer Tools: GitHub, Jira, Linear, Notion, Confluence, Figma Business Apps: Salesforce, HubSpot, Asana, ClickUp, Motion Communication: Gmail, Slack, Discord, Microsoft Teams Data & Analytics: PostgreSQL, Supabase, Google Sheets Content: YouTube, WordPress, document conversion tools
Each server handles the specific authentication method (OAuth2, API keys, etc.) and provides a consistent interface for your agents.
Klavis integrates with whatever you're already using:
You don't need to change your architecture or migrate to a new framework.
Start with the hosted API for rapid development:
pip install klavis
# Get API key from klavis.ai
# Start building immediately
Move to self-hosted when you need full control:
git clone https://github.com/klavis-ai/klavis.git
cd klavis/mcp_servers/github
docker build -t klavis-github .
docker run -p 8000:8000 klavis-github
All servers are open-source MIT licensed. You can inspect the code, modify it, or contribute improvements back to the community.
Klavis handles the operational concerns you don't want to think about:
pip install klavis or npm install klavisThe documentation at docs.klavis.ai includes working examples for every integration, OAuth setup guides, and multi-tool workflow patterns.
Your next AI agent project doesn't need to start with weeks of integration work. Start building the intelligence, not the plumbing.