Frequently asked.

Yes — that's the only way Znote uses AI. You connect your own provider (OpenAI, OpenRouter, Ollama…) and Znote never sits between you and it.

It depends on the provider you connect. With an OpenAI or OpenRouter key, queries go directly to that provider. For fully local AI, connect Ollama (LLaMA, Mistral…) on your own machine and nothing ever leaves it.

No. The app is free and there is nothing to buy from us. You pay your AI provider for what you consume — or nothing at all with Ollama.

Only what your provider charges, with no markup from us — a few euros a month covers daily use on most models. With Ollama running locally, it costs nothing.

MCP (Model Context Protocol) is the open standard that lets AI use real tools. Declare servers in your vault — GitHub, Postgres, Slack, Linear, your filesystem, hundreds more — and your AI blocks can query, search and create through them with tools=<server>. The config format is the same as Claude Desktop, so any server's README config pastes as-is.

Yes — each server is a local process that Znote starts and stops on your machine, consistent with everything else in the app. Your credentials live in your vault's .znote/mcp.json and are only used by the servers you configure, with your own tokens and scopes. Nothing goes through a Znote cloud.

Most servers run with npx (Node.js) or uvx (Python's uv) — if a runtime is missing, Znote shows the exact install command for your OS. Then start from the 22 ready-to-run examples: each one has the config and prompts to run as-is.