Cursor AI

Set up Cursor for your project

Cursor is one of the most popular AI-powered code editors, but it generates better code when it understands your project. Without a .cursorrules file and MCP server configs, Cursor falls back to generic suggestions. Caliber generates both automatically by scanning your codebase.

What Cursor needs to work well

Cursor reads two key configuration files:

• .cursorrules — a file in your project root that tells Cursor your coding conventions, preferred patterns, architecture, naming rules, and what to avoid. This is the most impactful file for Cursor's code quality. Learn more about what goes in a .cursorrules file
• .cursor/mcp.json — MCP server configurations that extend Cursor with external tools like database access, API documentation, and deployment capabilities. See the MCP setup guide

Manual Cursor setup vs Caliber

Setting up Cursor manually means writing a .cursorrules file from scratch — documenting every convention, pattern, and preference by hand. Most developers either skip it entirely or write a few lines that barely scratch the surface. MCP configuration is even more tedious: finding the right servers, getting the JSON format correct, and knowing which tools are relevant to your stack.

Caliber eliminates this entirely. It analyzes your actual codebase — your file structure, dependencies, naming conventions, test patterns, and import style — and generates a comprehensive .cursorrules file that reflects how your team actually writes code.

Set up Cursor with Caliber

1. 1Install Caliber — run npm install -g @rely-ai/caliber
2. 2Run init — navigate to your project and run caliber init. Caliber detects your framework, language, testing setup, and code patterns
3. 3Review output — Caliber generates .cursorrules with project-specific conventions and .cursor/mcp.json with recommended MCP servers
4. 4Keep it fresh — run caliber refresh to update .cursorrules as your project evolves

What Caliber detects for Cursor

• Naming conventions — camelCase vs snake_case, component naming, file naming patterns
• Framework patterns — React hooks, Next.js App Router, Express middleware, Django views
• Import style — absolute vs relative, path aliases, barrel exports
• Testing approach — test framework, file location, naming conventions, mocking patterns
• Error handling — error boundaries, try/catch patterns, custom error types

Caliber generates configs for every major AI tool from the same scan. Also set up Claude Code or configure all your AI tools at once.

Frequently asked questions