What are AI agents?

An AI agent configuration is a set of instructions that shapes how an AI assistant behaves in your project. Think of it as a personality profile plus a job description for your AI.

Out of the box, AI assistants are general-purpose. They’re helpful but generic. By configuring an agent with project-specific instructions, you get:

• Consistent code style — The AI follows your team’s conventions
• Domain knowledge — The AI understands your project’s context
• Better suggestions — More relevant code completions and advice
• Team alignment — Everyone’s AI assistant works the same way

A typical agent configuration includes:

Reusable instruction sets that teach the AI specific competencies. For example:

• “Write TypeScript with strict mode”
• “Follow conventional commit messages”
• “Use React Server Components where possible”

Guidelines and constraints for behavior:

• “Never use any type in TypeScript”
• “Always write tests for new functions”
• “Prefer composition over inheritance”

Background information about your project:

• Architecture decisions
• Key dependencies
• Business domain concepts

Different AI tools expect configurations in different formats:

Tool	Format	Location
Claude Code	Markdown	CLAUDE.md
Cursor	MDC	.cursor/rules/*.mdc
GitHub Copilot	Markdown	.github/copilot-instructions.md

spaget solves this by letting you build one agent configuration and export it to any platform.

• Getting started — Build your first agent in under 5 minutes
• Skills explained — Deep dive into the skill system