Varyz for Cursor

• Your preferences on demand. When the Cursor agent is about to assume something about you, such as a colour, a word, a tool or what to leave out, it asks Varyz first. If you already decided, it gets your answer.
• A briefing for the task. At the start of a task the agent can ask Varyz how you work and what this project has already settled.
• Only new questions reach you. When a preference or fact is genuinely missing, the questions come together as short choices.
• A finish check. Before the agent reports that it is done, it can ask Varyz for your checklist for that kind of work.
• The same rules as your other agents. What you settle while working in Claude Code or Codex also applies in Cursor, so your agents stay consistent.

Varyz runs on your own computer. Connecting it to Cursor needs no account and no API key.

• A local MCP server. Varyz is added to Cursor's MCP configuration and runs over stdio on your machine.
• A skill. It tells the Cursor agent when to consult Varyz: whenever it is about to assume something about you.

Cursor connects through MCP and the skill alone, without hooks. The agent calls Varyz itself when the skill applies, so the briefing and the finish check depend on the agent following it.

Varyz learns how you work from your past messages to Claude Code and Codex. It does not read Cursor chat history. If you work only in Cursor, Varyz learns from the answers you give when it asks, and from rules you confirm. You can review, replace or take back anything it learned.

Keep project conventions in Cursor rules, where they apply to everyone who opens the project. Varyz carries what is yours: your taste, your decisions and your standards, across every project and every agent. How Varyz compares with CLAUDE.md, AGENTS.md and Cursor rules.

Download local Varyz for free on Windows, macOS or Linux. Also see Varyz for Claude Code, Varyz for Codex and how Varyz works.