Vorec — AI Agent Skill for Screen Recording & Narrated Video (Codex, Claude Code, Cursor, Gemini CLI, Copilot)

Tell your agent what to record. The Vorec Recorder for macOS captures it. AI writes and voices the narration. Works with Claude Code (ready-made plugin) and any agent that can run terminal commands: Codex, Cursor, Gemini CLI, GitHub Copilot, Windsurf, OpenCode and more.

Your agent opens the browser and records anything you can click through: a product demo of the app it just built, a how-to for any website (Google Docs, Canva, Notion, Shopify admin, your own SaaS), a bug reproduction, a QA run, an onboarding tour or a release video. For example, a product demo: it drives your real running app, saves the MP4 to your disk for review, then AI writes and voices narration synced to every click. Claude Code uses the plugin in this repo; every other agent uses the same workflow through @vorec/cli and the agent skill at vorec.ai/skill.md.

More examples, with prompts and what an agent should not record: What can an AI agent record? 12 examples.

Each one comes back as a narrated MP4 (AI voice-over, zoom, cursor polish, subtitles, up to 4K) plus an optional written step-by-step guide.

Vorec doesn't depend on the model: anything that can run npx @vorec/cli@latest ... in a terminal can plan the flow, record it and hand you the MP4.

You: "record a tutorial showing how to sign up on vorec.ai"
  ↓
Claude writes a vorec.json manifest
  ↓
The Vorec Recorder for macOS captures the screen + clicks + cursor
  ↓
You review the MP4 locally
  ↓
Vorec uploads + generates AI narration + voice-over
  ↓
You polish in the Vorec editor (zoom, callouts, subtitles, music)
  ↓
Export HD video

No screen recording software to learn. No microphone. No video editor. Just one command in Claude Code.

Download the desktop app from vorec.ai/download.

After installing:

1. Open Vorec Recorder, sign in with your Vorec account
2. Grant Screen Recording and Accessibility permissions when prompted
3. Keep the app running (it lives in the menu bar)

macOS only. The recorder uses native ScreenCaptureKit for high-fidelity capture and synthesizes a cursor overlay during export.

1. In Claude Code, run /plugin
2. Go to Marketplaces → Add Marketplace
3. Enter: MustaphaSteph/vorec-plugins
4. Select record-tutorial and install it
5. Enable auto-update for the marketplace

No plugin needed. Any agent that can run terminal commands drives the same CLI:

1. Install the Vorec Recorder (step 1) and brew install cliclick.
2. Save your API key once (from vorec.ai → Settings → API Keys): npx @vorec/cli@latest init.
3. Tell your agent how to use Vorec by adding this to its instructions file (AGENTS.md for Codex and most agents, .cursor/rules/vorec.md for Cursor, GEMINI.md for Gemini CLI, .github/copilot-instructions.md for Copilot):

## Recording demo videos
When asked to record a demo, tutorial or walkthrough of this app, use Vorec:
follow https://vorec.ai/skill.md. Write a vorec.json manifest, run
`npx @vorec/cli@latest run vorec.json`, show me the MP4, and only after I
approve run `npx @vorec/cli@latest analyze <mp4>` to add AI narration.

A copy-paste version lives in examples/AGENTS.md.

Get your API key from vorec.ai → Settings → API Keys, then:

npx @vorec/cli@latest init

Just tell Claude Code what to record:

"record a tutorial showing how to create a project"
"make a demo video of the settings page"
"create a screencast of the signup flow on vorec.ai"
"show how the onboarding experience works"
"demo the checkout flow for the team"

Claude writes a vorec.json manifest, the Vorec Recorder captures the screen, you review the MP4, and Vorec generates the narration. End-to-end in minutes.

Building an app is hard enough. Making tutorial videos for it shouldn't be.

Vorec records your screen and generates professional narrated videos — right from your terminal. No screen recording software to learn, no microphone, no video editing skills needed.

• Just say what to record — "show how to sign up", "demo the billing page", "walk through the dashboard"
• Native macOS capture — high-fidelity ScreenCaptureKit recording with click + cursor tracking
• AI narration — professional voice-over generated automatically from the action timeline
• Local-first review — see the raw MP4 before anything uploads, approve only what's worth narrating
• Full video editor — synthetic cursor with ripples, zoom + chain pan, callouts, subtitles, background music, HD export

You're building fast with AI. Your app works, but you need tutorials for users and docs for your README. Vorec creates them in minutes — right inside Claude Code where you're already working. No context switching.

Investors want to see the product, not slides. Record a polished product demo in minutes — with AI voice-over that explains every feature clearly. Ship your pitch deck with a video that sells.

You're doing everything yourself — building, marketing, support. You don't have time to record, narrate, and edit tutorial videos. Tell Vorec what to show, review the recording, get a finished video with narration.

Your product updates faster than your docs. Every feature release needs a walkthrough, every changelog needs a demo. Vorec turns "record the new dashboard" into a narrated video your support team can share.

Create onboarding tutorials, API walkthroughs, and getting-started guides without recording yourself. Vorec generates natural narration that explains each step — in multiple languages.

Customer onboarding is where you lose users. Turn your onboarding flow into a professional tutorial video — with zoom effects highlighting important buttons, callouts explaining features, and subtitles for accessibility.

Your project needs a "Getting Started" video but you hate being on camera. Vorec records the setup flow and narrates it for you. Embed it in your README and watch contributions grow.

Delivering a project? Include a narrated walkthrough video for your client. Takes 5 minutes. Looks like you spent hours. Clients love it.

Write tutorials about tools you use. Vorec records the flow, generates narration, and turns it into a polished video you can embed in blog posts, YouTube, or social.

• Native macOS capture via ScreenCaptureKit (high-fidelity, no browser sandbox)
• Every action tracked — clicks, keystrokes, scrolls, cursor path, timing
• Local MP4 — review before anything uploads, keep or discard
• Cursor track sidecar — feeds the synthetic-cursor renderer for perfect frame parity in editor + export

• AI voice-over — natural narration that explains each step
• Synthetic cursor + click ripples — animated cursor with shape-aware sprites (arrow, pointing, text, resize, etc.) and click ripples on every interaction
• Zoom & spotlight — auto-zoom from cursor dwell + clicks, chain-pan between zoom regions, manual zoom clips
• Callouts & shapes — arrows, circles, boxes, number badges
• Backgrounds — gradients, wallpapers, rounded corners, shadows
• Intro slides — title cards with professional themes
• Background music — built-in library + your own uploads, volume + fade controls
• Subtitles — auto-generated, customizable, per-language
• Multi-language — translate narration to any language
• Timeline editor — adjust timing, re-record segments, trim, ripple-edit
• Export — HD MP4 via the Cloud Run export worker

Describe what you want and get a video. No scripts to write, no buttons to click, no clips to manage.

The Vorec Recorder writes the MP4 to your disk. You review it before anything uploads. Approve only the recordings you want narrated.

Every click, keystroke, scroll, and cursor pause is timestamped and tracked. The narration AI gets a full action timeline — no guessing from frames.

Record once, narrate in any language. Translations are stored per-segment and overlaid by the export worker at render time.

• macOS (Apple Silicon or Intel)
• Vorec Recorder — download from vorec.ai/download, sign in, grant Screen Recording + Accessibility
• Node.js 18+ for the CLI
• cliclick (for action injection during automated recording):

  brew install cliclick

• Vorec API key — get from vorec.ai → Settings → API Keys

The plugin runs npx @vorec/cli check before recording to verify all four are in place.

npx @vorec/cli@latest init                    # Save API key to ~/.vorec/config.json
npx @vorec/cli@latest check                   # Verify app + sign-in + permissions + cliclick
npx @vorec/cli@latest run <vorec.json>        # Record locally — writes MP4 + sidecar, no upload
npx @vorec/cli@latest analyze <video.mp4>     # Upload + generate narration
npx @vorec/cli@latest projects list --json    # List accessible projects
npx @vorec/cli@latest status                  # Check processing status
npx @vorec/cli@latest editor inspect --project <id> --json
npx @vorec/cli@latest editor snapshot --project <id> --at 42.5 --output frame.png
npx @vorec/cli@latest editor filmstrip --project <id> --every 2 --output-dir frames
npx @vorec/cli@latest editor describe --project <id>
npx @vorec/cli@latest actions list --project <id> --json
npx @vorec/cli@latest actions move --project <id> --click-index 7 --at 46.4
npx @vorec/cli@latest actions update --project <id> --click-index 7 --x 386 --y 381
npx @vorec/cli@latest actions set-primary --project <id> --segment <segment-id> --click-index 7
npx @vorec/cli@latest actions verify --project <id> --click-index 7 --output action-7.png
npx @vorec/cli@latest narration move --project <id> --segment <segment-id> --at 46.4
npx @vorec/cli@latest narration update --project <id> --segment <segment-id> --text "Now save the settings."
npx @vorec/cli@latest narration attach-action --project <id> --segment <segment-id> --click-index 7 --primary
npx @vorec/cli@latest overlays list --project <id> --json
npx @vorec/cli@latest overlays add --project <id> --type zoom --at 12 --duration 3 --x 430 --y 360
npx @vorec/cli@latest overlays add --project <id> --type follow-zoom --at 18 --duration 4
npx @vorec/cli@latest overlays add --project <id> --type blur --at 22 --duration 2 --x 700 --y 180 --width 220 --height 90
npx @vorec/cli@latest overlays add-image logo.png --project <id> --at 0 --duration 4
npx @vorec/cli@latest overlays move --project <id> --clip <clip-id> --at 14.2 --x 480 --y 340
npx @vorec/cli@latest overlays resize --project <id> --clip <clip-id> --duration 4 --width 280 --height 160
npx @vorec/cli@latest overlays update --project <id> --clip <clip-id> --data '{"compositionState":{"keyframes":[{"property":"destination.x","clock":"story","keyframes":[{"id":"kf-1","offsetSeconds":0,"value":400},{"id":"kf-2","offsetSeconds":1.5,"value":600}]}]}}'
npx @vorec/cli@latest overlays delete --project <id> --clip <clip-id>
npx @vorec/cli@latest cursor hide --project <id>
npx @vorec/cli@latest cursor show --project <id>
npx @vorec/cli@latest background get --project <id> --json
npx @vorec/cli@latest background set --project <id> --type gradient --gradient-start "#0f172a" --gradient-end "#4f46e5" --angle 135 --padding 8 --shadow large
npx @vorec/cli@latest background set --project <id> --type wallpaper --wallpaper ocean
npx @vorec/cli@latest background disable --project <id>
npx @vorec/cli@latest segments --project <id> # Read narration segments (JSON)
npx @vorec/cli@latest media upload intro.mp4 --project <id> --wait
npx @vorec/cli@latest media list --project <id>
npx @vorec/cli@latest timeline list --project <id>
npx @vorec/cli@latest timeline split --project <id> --at 42.5
npx @vorec/cli@latest timeline add-video intro.mp4 --project <id> --position intro --muted
npx @vorec/cli@latest timeline add-video broll.mp4 --project <id> --after-segment <video-segment-id> --muted
npx @vorec/cli@latest export start --project <id> --wait --download-url
npx @vorec/cli@latest export status <export-id> --download-url
npx @vorec/cli@latest export cancel <export-id>

The run and analyze split lets you review the raw MP4 before paying credits to analyze it. The agent always shows you the recording first.

Editor media/timeline/action/overlay/export commands are optional post-analysis controls. Use them only when the user asks to inspect, edit, or export an existing Vorec project. Use editor snapshot / editor filmstrip for visual readback. Use editor describe only when Vorec AI should analyze the source video; it spends analysis credits. For exact mid-segment insertion, split first, then add the media at the split timestamp.

Narration segments reference actions by index. In editor inspect --json, narrationSegments[].primaryClickIndex and clickRefs[] match clicks[].clickIndex. The clicks[] entries are the source of truth for action timestampSeconds, x, y, and interactionType; segment coordinates are not the source of truth. Use narration move/update/attach-action for voice segment edits, and actions verify to render the frame at an action timestamp.

Overlay clips are normal editor timeline clips. Agents can add/update/delete zoom, follow-zoom, blur, spotlight, callout, text, shape, image, slide, and cursor clips, upload local image overlays with overlays add-image, then move/resize them with normalized 0..1000 video coordinates. Cursor visibility is controlled separately with cursor show|hide. Video background is a project setting controlled with background get/set/disable; inspect it structurally before changing it. Cursor visibility and background changes now create undoable timeline revisions.

Advanced overlays add/update --data payloads can author canonical keyframes and animation applications. Vorec validates animation payloads before writing them. If the CLI reports (field=...), fix that exact field; do not remove animation data or retry the same malformed payload. Inspect the overlay again after the mutation so unrelated clip data and authored animation remain intact.

Every export start is pinned to one immutable canonical composition revision. The CLI prints the pinned revision and content hash (and returns compositionRevision / compositionContentHash with --json), so later edits cannot change an in-flight render. Only the project owner can start an export. If no canonical revision is available, inspect and save the project instead of falling back to stale state.

Video track segments are visible through timeline list --json and editor inspect --json. Agents should use videoSegments[].id as insertion anchors when possible: after splitting a source segment, insert with timeline add-video --after-segment <firstSegment.id> or --before-segment <secondSegment.id> instead of retyping the split timestamp.

This plugin records on macOS only via the Vorec Recorder desktop app. The recorder uses Apple's ScreenCaptureKit framework, which has no equivalent on Windows or Linux.

The narration and editing side (vorec.ai web app) works in any modern browser.

node scripts/validate-plugin.mjs

This verifies plugin JSON, tracked-action sample data, internal markdown links, stale setup claims, script syntax, root licensing, and accidental .DS_Store tracking.

See docs/release-checklist.md for the full release checklist and examples/common-flows.md for common recording patterns.

• vorec.ai — AI-narrated tutorial videos
• vorec.ai/download — Vorec Recorder for macOS
• @vorec/cli on npm — CLI tool
• Plugin Marketplace — This repo

Is there a plugin that lets an AI agent record my screen? Yes: the record-tutorial plugin for Claude Code in this repo, and the same workflow for every other agent through the Vorec CLI.

Is there an agent skill for screen recording or making videos? Yes. skills/record-tutorial/SKILL.md is a standard agent skill: it tells the agent how to plan a recording, write the manifest, drive the recorder and hand you the MP4.

Is there an MCP server for screen recording with Vorec? Not yet. Agents use the CLI instead (npx @vorec/cli@latest ...), which works in any agent that can run terminal commands, including Codex, Cursor, Gemini CLI, Copilot and Claude Code.

Is there a ChatGPT or Claude connector for Vorec? Not yet. In chat apps without a terminal, upload your screen recording at vorec.ai and Vorec narrates it.

Is there a Codex skill (or Cursor, Gemini CLI, Copilot skill) for recording videos? Yes, this one. Every agent that can run terminal commands uses the same Vorec CLI; paste the snippet from examples/AGENTS.md into AGENTS.md (Codex), .cursor/rules (Cursor), GEMINI.md (Gemini CLI) or .github/copilot-instructions.md (Copilot). Agents that load SKILL.md-style agent skills can use skills/record-tutorial directly.

Can an AI agent make videos? Yes. The agent plans the walkthrough, drives your real app or any website through the Vorec Recorder on macOS, and saves an MP4. Vorec then writes and voices the narration and adds zoom, cursor polish and subtitles.

Can Claude Code make videos? Yes. With this plugin, Claude Code plans the walkthrough, drives your real app through the Vorec Recorder on macOS, and saves an MP4. Vorec then writes and voices the narration and adds zoom, cursor polish and subtitles.

Can Claude Code edit videos? Yes, Vorec projects. The CLI lets the agent inspect the timeline, move or rewrite narration, add zoom, blur, callouts, text and image overlays, change the background, split the timeline, add intro/outro clips and start the export (see CLI Commands).

Can Codex, Cursor or Gemini CLI record my screen? Yes. Any agent that can run terminal commands uses the same CLI: npx @vorec/cli@latest run vorec.json. Add the snippet from examples/AGENTS.md to your agent's instructions file.

What can I record? Product demos, how-tos on any website, onboarding tours, bug reproductions, QA runs, release videos, support answers, course lessons and sales tours. See 12 examples with prompts.

How do I install a Claude Code video plugin? In Claude Code run /plugin, add the marketplace MustaphaSteph/vorec-plugins, then install record-tutorial. Full steps in Quick Start.

Does it need a Mac? Recording does (the Vorec Recorder uses macOS ScreenCaptureKit). Already have a recording from Windows, Linux, Zoom or Loom? Upload it at vorec.ai for AI narration.

Is it free? The plugin and CLI are free and MIT-licensed. Narration runs on a Vorec account: a 7-day trial with 100 credits and no card, then plans from $9/month (pricing).

• What can an AI agent record? 12 examples with prompts
• Can an AI coding agent record your app demo? — Playwright, Remotion, agentic-demo and Vorec compared
• Let Claude Code record a demo of the app it built
• Record app demos with Codex
• One recording, two outputs: tutorial video + written guide
• Add AI narration to any screen recording
• Recording a meeting first? Zoom · Google Meet · Microsoft Teams · Loom (add a voiceover)

• Claude Code — AI coding agent by Anthropic
• Best Claude Code Skills & Plugins (2026)

Keywords: is there a plugin to record screen with ai, ai agent plugin screen recording, agent skill video, screen recording mcp, mcp server video recording, chatgpt connector video, claude code mcp, cursor mcp, codex mcp, gemini cli extension, agent skills, ai agent skills, codex skills, best codex skills, cursor agent skills, cursor rules, gemini cli extensions, github copilot agent skills, opencode agent skills, awesome agent skills, agent skills github, claude code video, claude code video plugin, claude code video editing, claude code skills, claude code plugins, best claude code plugins, claude code skills github, agent skill screen recording, how to make a tutorial video, record a course with ai, record zoom meeting, record google meet, record teams meeting, record anything with an ai agent, ai agent browser recording, record how-to video with ai, bug reproduction video, qa recording, claude opus 5.5, gpt-6.1 sol, gemini 4 argon, google antigravity, ai agent screen recording, codex demo video, cursor demo video, gemini cli, github copilot, ai coding agent demo recorder, claude code plugin, claude code skill, screen recording, AI narration, tutorial video generator, product demo recorder, investor demo, screencast tool, automated documentation, voice-over generator, video tutorial maker, SaaS demo, developer onboarding, vibe coding, solo developer tools, startup demo video, Claude Code screen recorder, AI tutorial maker, product walkthrough, open source documentation, customer onboarding video, changelog demo, training video maker, macOS screen recorder, ScreenCaptureKit recorder, Vorec Recorder, content creator tools