Agent Support Matrix¶
This page documents feature availability across the four supported AI coding agents. Use it as a quick reference when you encounter agent-specific instructions throughout the workshop.
Feature Availability¶
| Feature | Claude Code | Kiro | Cursor | Codex |
|---|---|---|---|---|
| Plugin install | ✅ | ✅ | ❌¹ | ✅ |
| MCP Server | ✅ | ✅ | ✅ | ✅ |
| Skills discovery | ✅ | ✅ | ✅ | ✅ |
| Rules file | ✅ CLAUDE.md |
✅ .kiro/steering/ |
✅ .cursorrules |
✅ codex.md |
| run_script sandbox | ✅ | ✅ | ✅ | ✅ |
¹ Cursor: Manual MCP JSON configuration required instead of plugin install. See the Cursor tab in Agent Setup for the exact
mcp.jsonformat.
What Each Feature Means¶
- Plugin install — One-command setup that registers the MCP Server endpoint and configures credentials automatically.
- MCP Server — The agent can call AWS APIs through the Agent Toolkit MCP endpoint at
https://mcp.us-east-1.amazonaws.com/. - Skills discovery — The agent can browse and invoke pre-built AWS skill prompts for common tasks (deploying resources, writing IaC, debugging).
- Rules file — A project-level configuration file that sets guardrails, security boundaries, and behavioral constraints for the agent.
- run_script sandbox — The agent can execute Python/bash scripts in a sandboxed environment for validation and automation tasks.
Impact on Workshop Labs¶
All core labs work with every agent — the main instruction body uses natural-language prompts that are agent-agnostic. Where a feature is unavailable (marked ❌), the relevant lab tab provides the workaround so you can achieve the same outcome with your chosen agent.