Multi-Runtime Support

CAP Pro is the only Code-First framework that runs across 8 AI coding runtimes from a single install:

• Claude Code
• Gemini CLI
• Codex
• GitHub Copilot
• Cursor
• Windsurf
• OpenCode
• Antigravity

The installer translates the canonical Claude Code agent/command format into each runtime's native format and writes them into the right config directory.

Why care?

Most frameworks lock you into one runtime. If you switch tools, you lose your workflow. CAP Pro's tags and Feature Map travel with the project, not the tool — so when you switch from Claude Code to Cursor or back, your CAP Pro state is intact.

What gets translated

Source (Claude Code)	Translated to
~/.claude/agents/cap-*.md (Markdown agent with YAML frontmatter)	Native agent format for each runtime
~/.claude/commands/cap/*.md (slash command)	Skill / command for each runtime
~/.claude/hooks/*	Native hook format where supported
Tool names (Read, Edit, Bash, Grep)	Mapped to runtime-specific tool names (e.g. Copilot's read, edit, execute, search)

Per-runtime install commands

# Pick what you use
npx cap-pro@latest --claude
npx cap-pro@latest --cursor
npx cap-pro@latest --windsurf

# Or install for everything at once
npx cap-pro@latest --all

Slash command syntax differences

Runtime	Syntax
Claude Code	/cap:prototype
Gemini CLI	/cap:prototype
OpenCode	/cap-prototype
Codex	$cap-prototype
GitHub Copilot	/cap-prototype
Antigravity	/cap-prototype
Cursor	cap-prototype (mention the skill name)
Windsurf	/cap-prototype

The CAP Pro installer handles all of this — you don't have to remember.

Tool-name mapping (Copilot example)

GitHub Copilot uses different tool names than Claude. The installer maps them:

Claude Code	Copilot
Read	read
Write, Edit	edit
Bash	execute
Grep, Glob	search
Task	agent
WebSearch, WebFetch	web
TodoWrite	todo
AskUserQuestion	ask_user
SlashCommand	skill

This mapping applies only to agents. Skills (slash commands) are passed through unchanged.

Custom config directories

Multi-config users (work vs personal, multiple Claude profiles):

npx cap-pro@latest --claude --global --config-dir ~/.claude-work

Takes priority over runtime-specific env vars (CLAUDE_CONFIG_DIR, GEMINI_CONFIG_DIR, CODEX_HOME, COPILOT_CONFIG_DIR, ANTIGRAVITY_CONFIG_DIR, CURSOR_CONFIG_DIR, WINDSURF_CONFIG_DIR).

What does NOT translate

A few features are runtime-specific:

• Statusline integration — Claude Code and Gemini CLI only.
• Session-extract subcommand (cap extract) — relies on Claude Code's session log format.
• Hooks — only Claude Code, OpenCode, Gemini and Codex have a stable hook system. Copilot and Cursor don't expose pre/post-tool hooks the same way.

For these, CAP Pro degrades gracefully — the affected feature is a no-op on unsupported runtimes, but everything else works.

Roadmap: more runtimes?

If a new AI coding runtime appears (Sourcegraph Cody, Aider, Continue, …) and has a stable agent + skill format, we'll add it. Open an issue with the runtime's spec link.