SDD with Different AI Agents

cc-sdd Supports 8+ AI Coding Agents

The same .kiro/ spec files work across every supported agent. You can start a feature with one agent and continue with another — the specs are agent-agnostic. Install with the appropriate flag to configure for your preferred tool.

Agent-by-Agent Guide

Claude Code — Primary target. Full agentic coding: reads and writes files across your entire codebase, runs terminal commands, runs tests, makes multi-file changes. Best overall SDD experience because Claude Code was designed for the kind of structured, multi-step execution the SDD pipeline requires.

Install: npx cc-sdd@latest (default — no flag needed)

Cursor — IDE-integrated. Composer handles multi-file edits. SDD specs appear in the project sidebar. Best for teams that want SDD integrated into their daily editing workflow. The Cursor Composer sees your .kiro/ specs automatically.

Install: npx cc-sdd@latest --cursor

GitHub Copilot — Workspace feature handles multi-file tasks. cc-sdd bundles prompt packs optimized for Copilot's context window. Best for teams already using GitHub Copilot enterprise.

Install: npx cc-sdd@latest --copilot

Codex CLI (OpenAI) — Terminal-based agent. Good for headless environments and CI/CD pipelines. Best when you want SDD in automated workflows.

Install: npx cc-sdd@latest --codex

Gemini CLI (Google) — Terminal agent with the longest context window of any agent. Exceptional for large codebases where you need the AI to read many files before generating output.

Install: npx cc-sdd@latest --gemini

Windsurf — AI-native IDE with the "Cascade" multi-step agent. Good parallel to Cursor. Best for teams adopting an AI-native development environment.

Install: npx cc-sdd@latest --windsurf

OpenCode — Open-source terminal agent. No API key vendor lock-in. Best for teams that want full control over their AI tooling.

Install: npx cc-sdd@latest --opencode

Qwen Code — Alibaba's coding agent. Strong for teams in regions with Alibaba Cloud infrastructure.

Install: npx cc-sdd@latest --qwen

Choosing the Right Agent

You are not locked in. The .kiro/ specs are agent-agnostic. Start a feature with Claude Code, hand off to a developer using Cursor — the specs work for both.

Agent Portability

Because all agents read the same .kiro/ spec files, teams can use different agents for different stages:

1. Use Gemini CLI for steering (reads your large codebase efficiently)
2. Use Claude Code for implementation (best multi-file autonomous coding)
3. Use Cursor for review and refinement (IDE-integrated, great for targeted edits)

Customizing Templates and Rules

Edit .kiro/settings/templates/ to customize the output format for requirements.md, design.md, and tasks.md to match your team's standards.

Edit .kiro/settings/rules/ for domain-specific standards that apply across all features (not just per-project steering files).

Multi-language support: cc-sdd generates spec files in 13 languages. Use the --lang flag:

npx cc-sdd@latest --lang ja  # Japanese
npx cc-sdd@latest --lang es  # Spanish
npx cc-sdd@latest --lang zh  # Chinese

Key Takeaways

• cc-sdd supports Claude Code, Cursor, GitHub Copilot, Codex, Gemini CLI, Windsurf, OpenCode, and Qwen
• .kiro/ spec files are agent-agnostic — switch agents mid-feature without losing your specs
• Claude Code gives the best overall SDD experience; Cursor and Windsurf integrate best into IDE workflows
• Gemini CLI has the largest context window — best for reading large codebases during steering
• Customize templates in .kiro/settings/templates/ and domain rules in .kiro/settings/rules/