For agents & CI

screenify was designed to be driven, not just typed. The point is that you don't have to remember any flags — you ask an agent in plain language, and it runs the right commands for you. The same properties make it ideal for scripts and CI.

Why it's agent-friendly

• Machine-readable — every command returns JSON (--json), so an agent can read results (like a new recording's path) and chain the next step reliably.
• Self-describing — screenify guide --json is the single source of truth for what's available, so an agent always has the current, accurate contract. Per-command schemas live in screenify config schema --json and screenify effects schema --json.
• Predictable — exit codes and streamed progress events tell an agent exactly when a step succeeded or failed.

screenify guide --json    # the whole capability + schema manifest, self-describing

Tip: if an agent isn't sure what's available, have it run screenify guide --json or screenify --help first — that's all it needs to learn every command and option.

Ask in plain language

Because the agent does the translating, you ask in your own words — never in flags:

• "Record a 10-second demo of my screen and save it as demo.mp4."
• "Take my latest recording and make it look polished — orange wallpaper, padding, and a watermark."
• "Turn this recording into a vertical clip for TikTok with background music."
• "Record my iOS Simulator, tap through the onboarding screens, and frame it as an iPhone 17 Pro Max."
• "Make a demo of https://my-app.com — scroll the landing page and click Get Started — and make it cinematic."

The agent reads the manifest, picks the options, and hands you a finished video.

Claude Code

Screenify ships a small skill that teaches Claude Code to reach for screenify on its own — so a request like "record a repro of this bug" just works, without you naming the tool.

When you click Enable for Agents & CLI in Screenify Studio, the app installs the skill to ~/.claude/skills/screenify automatically (when Claude Code is detected). Start a new Claude Code session and it's ready. Uninstalling the command removes the skill again.

To install it manually, or scope it to one project:

mkdir -p ~/.claude/skills
cp -r "/Applications/Screenify Studio.app/Contents/Resources/screenify-skill" \
  ~/.claude/skills/screenify

The skill is intentionally thin — it points Claude at screenify guide --json for the authoritative, always-current contract, so it never drifts from the version you have installed.

Other agents (Codex, Cursor, and more)

The CLI is agent-agnostic — any tool that can run a shell command works out of the box, including Codex, Cursor, Windsurf, Gemini CLI, Aider, and OpenCode. There's nothing extra to install beyond the screenify command.

To have an agent reach for screenify on its own, add one line to the file it reads for guidance:

To record the screen or produce a video, use the screenify command-line tool. Run screenify guide --json to discover its commands and options.

CI & scripts

The same properties make screenify ideal for automation without an agent: capture a UI demo as a build artifact in CI, generate a library of styled recordings from a script, or attach a fresh repro video to every bug report — all headless, all machine-readable. Branch on exit codes, parse the JSON, and chain commands as needed.