Why we build Klavis AI:

While working with various AI models and external tools, we think MCPs are useful but you need to install and manage them manually. Also there is no tool for integrating them into team workflows on platforms like Slack and Discord.

What is it:

Klavis provides hosted, Open Source MCP clients on Slack/Discord and hosted various MCP servers. It consists of:

- MCP Clients (in the mcp_clients/ directory): We currently have clients for Slack and Discord bot.

- MCP Servers (in the mcp_servers/ directory): These are standalone MCP servers that wrap specific tools or capabilities, exposing them via a standardized MCP interface (specifically, Server-Sent Events).

For MCP servers, we have built:

- Klavis ReportGen (LLM + Web Search): This is a very cool one. Check out the report generated by the server with a simple query “generate a report about national parks in utah” https://www.klavis.ai/generated-reports/c150d577-ab66-45e4-9...

- YouTube video

- Document Conversion (via Pandoc)

- And we modified various servers to support SSE.

How to Try It Out:

- We have our hosted version in www.klavis.ai, which has a UI to directly configure MCP clients and servers.

- The github repo also has detailed instructions for local development and testing. The basic steps involve setting up a client (Discord or Slack) and running one or more MCP servers (like Report Generation). You can check the readme under mcp_clients and mcp_servers folder for details.

- For those of you who do not have much time, check out our launch video: https://www.youtube.com/watch?v=9-QQAhrQWw8

This is an early release, and I'd appreciate feedback from the HN community:

- What other tool servers (MCP endpoints) or client platforms would be most valuable?

- Any thoughts on improving the developer experience for adding new servers/clients?

- Discord for feedback: https://discord.gg/P6fFgv2w

We'll be around to answer questions. Thanks for looking!