# Kondi MCP server

> Build multi-LLM workflow pipelines using deliberating councils of specialized AI agents that execute tasks through MCP-connected tools.

- Repository: https://github.com/thisPointOn/kondi
- Page: https://hostedmcp.org/servers/thispointon/kondi
- Transport: —
- Stars: 5
- License: MIT

## Tools (4)

- `read_file`
- `write_file`
- `list_directory`
- `run_command`

## FAQ

### How does Kondi handle MCP server integration?

You can connect to remote servers via SSE/HTTP, local servers via stdio, use the built-in library, or install directly from a GitHub repository URL.

### Can I use different AI providers within a single workflow?

Yes, each persona in a council can use a different provider and model, including Anthropic, OpenAI, DeepSeek, Google, xAI, and more.

### Does Kondi require an account or subscription fees?

No, Kondi is a free, open-source desktop app that requires no account and no fees; you bring your own API keys.

### How are OAuth-protected MCP servers handled?

Kondi spawns a local Node.js proxy per server to manage PKCE browser flows, token refresh, and retry logic.
