klavis-ai/klavis

Klavis is a platform that helps AI agents connect to and use external tools reliably. The core concept it builds around is called MCP, short for Model Context Protocol, which is a standard way for AI systems to call out to services like Gmail, Slack, GitHub, or hundreds of others. Rather than building and maintaining those connections yourself, Klavis provides them ready to use with built-in authentication support. The platform has three main pieces. The first is called Strata, which acts as a smart connector between an AI agent and a large collection of tools. When an agent has access to many tools at once, the list of available actions can become too long for the model to handle well. Strata is designed to manage that by surfacing only the relevant tools in a given moment, keeping the model's working memory focused. The second piece is a library of more than 100 prebuilt MCP integrations covering popular services, each with OAuth login support so users can authenticate without developers handling credentials manually. The third piece is a Sandbox environment intended for teams training AI models, providing a controlled space where language models can practice using tools at scale. You can use Klavis through their hosted cloud service, run it yourself with Docker, call it from Python or TypeScript code using their SDK, or use plain HTTP requests through a REST API. A short code example shows creating a connection to Gmail and Slack in a few lines of Python. The project is open source and maintained by the Klavis team. Documentation, a community Discord, and a link to report issues are provided on their website.