clipstitchr
v1.4.5
Published
Record demos, start new content, and queue finished Stitches in ClipStitchr from your terminal.
Maintainers
Readme
ClipStitchr CLI
Record guided product demos from a local or live app, upload finished demo files, start new content, and queue ready content in ClipStitchr.
npx clipstitchrRunning clipstitchr opens one persistent terminal workspace. Its header shows
the current product, repo connection, and account connection from local CLI
settings. Missing account or repo setup appears first; after setup, the main
menu keeps everyday creation and queue actions first. The full-screen workspace
keeps the header at the top instead of adding new frames to terminal history.
Choose Demos, then pick Record it myself for manual recording or Let AI
record it for me for a policy-guarded AI run with a saved guide.
Slash commands let you stay in the interactive shell while running direct
commands, such as /demo manual, /demo agent --guide "Checkout flow",
/queue stitch --all, /products use product_123, and /status. The slash
composer ranks command names, subcommands, option names, and meaningful command
tokens as you type. For example, /policy edit finds /demo policy edit, and
/queue all finds the queue-all choices. One nearby typo is tolerated for
longer command tokens. Press Tab to complete the highlighted suggestion. Enter
runs a complete command or expands a command group so you can keep typing.
Ctrl+P or Ctrl+N moves through command history, and Escape returns to the
current menu. Suggestions and header context are local and deterministic; they
do not use AI or call the ClipStitchr server.
After an action finishes, its output stays in a result view instead of being
replaced by the action menu. Use Up/Down or j/k to scroll longer results.
Home/End and Page Up/Page Down move faster when those keys are available. Use
Tab, Shift+Tab, or Left/Right to choose Back, Main menu, Type a slash command,
or Exit, then press Enter. Menu height also follows the terminal height so the
ClipStitchr header stays visible in shorter windows. Result pages reserve room
for several lines before scrolling on a typical terminal.
Menu mode uses one compact shortcut line instead of a separate prompt and
bordered shortcut box, leaving room for several choices in shorter terminals.
Run clipstitchr queue when you want a focused queue menu for latest, all, or
specific Stitch and Swipe queue actions.
Useful commands:
clipstitchr help
clipstitchr --version
clipstitchr link
clipstitchr status
clipstitchr update
clipstitchr demo
clipstitchr demo agent
clipstitchr demo agent --guide "Checkout flow"
clipstitchr demo agent --driver openai-computer
clipstitchr demo agent --driver openai-computer --openai-mode relay
clipstitchr demo agent --driver openai-computer --target live --url https://example.com
clipstitchr demo agent --driver openai-computer --surface macos-window --openai-mode relay
clipstitchr demo guide create
clipstitchr demo guide list
clipstitchr demo guide show "Checkout flow"
clipstitchr demo guide edit "Checkout flow"
clipstitchr demo guide delete guide_123
clipstitchr demo guide save-instructions guide_123
clipstitchr demo policy init
clipstitchr demo policy check
clipstitchr demo policy edit
clipstitchr demo agent --guide guide_123 --driver openai-computer
clipstitchr demo agent --guide guide_123 --driver openai-computer --target live --url https://example.com
clipstitchr demo agent --guide guide_123 --no-upload
clipstitchr demo logs agent_run_123
clipstitchr demo manual
clipstitchr demo manual --guide "Checkout flow"
clipstitchr demo manual --no-guide
clipstitchr demo upload ./demo.mp4
clipstitchr stitchr new
clipstitchr swipr new
clipstitchr queue
clipstitchr queue list
clipstitchr queue stitch
clipstitchr queue stitch --all
clipstitchr queue swipe
clipstitchr queue swipe --all
clipstitchr queue --all
clipstitchr products
clipstitchr products list
clipstitchr products create --use
clipstitchr products use
clipstitchr native init
clipstitchr native init --force
clipstitchr native check
clipstitchr unlink
clipstitchr --plain statusIf the repo is linked, your ClipStitchr account is connected, and the saved
browser profile is already signed into your app, clipstitchr demo agent
writes a guide with ClipStitchr AI and records the demo with the guarded AI
agent in one command. Use clipstitchr demo agent --guide "Checkout flow" when
you want to record from an existing guide. It can use localhost by default, or
a live/staging URL with --target live. It saves the guide, MP4, screenshots,
action log, and run summary locally. By default it asks you to review before
upload. Use --no-upload to keep the MP4 local or --upload to upload without
the review prompt.
The built-in recorder can also create a simple walkthrough checklist before
each demo. Run clipstitchr demo when you want a focused menu for recording,
AI recording, guides, safety policy setup, uploads, and logs. You can run
clipstitchr demo guide create directly to draft a guide with
ClipStitchr first, review it, edit it, and save it for the next recording.
Saved guides get readable names like Checkout flow, and demo guide list
shows those names first. You can use a guide name, ID, or file path with
demo guide show, edit, delete, save-instructions, demo manual
--guide, and demo agent --guide.
During recording, the terminal walks through each step and records section
timing metadata for ClipStitchr to use later for chapters, captions, smart
zooms, and editing decisions. Use --no-guide when you want one free-form take,
or --guide when you want to reuse a saved guide from
.clipstitchr/demo-guides.
Recording is still manual by default: the CLI opens your app in Chromium, you
click through the demo, then press Enter in the terminal as each step is done.
If your app requires login, sign in inside the recorder browser once. The CLI
keeps that app browser session in .clipstitchr/browser-profile so future
recordings can stay signed in.
The demo agent beta is policy guarded. clipstitchr demo policy init creates
the local safety settings and lets you review them before saving.
clipstitchr demo policy check confirms the saved policy is valid. Localhost
app URLs are the default. Live or staging URLs need a separate yes before they
are allowed. File uploads stay off unless you name the exact local files the
agent may use. If you turn uploads on but leave the file list empty, uploads
stay off and the rest of the policy is still saved. Run clipstitchr demo policy edit when routes, test values,
blocked words, upload files, or time limits need to change. clipstitchr demo
auto creates the same safe policy automatically when one does not exist.
The automatic agent uses OpenAI Computer Use when OPENAI_API_KEY is available
locally, or the hosted ClipStitchr relay when you are logged in and no local key
is available. Relay mode sends screenshots through ClipStitchr servers and
never sends a server OpenAI key back to the CLI. If neither direct nor relay
mode is available, it falls back to the structured planner. clipstitchr demo
agent init and clipstitchr demo agent check still work as legacy aliases,
but new setup should use demo policy. The same policy validator always
decides what can run.
Most demos work best around 30-90 seconds. Longer recordings are allowed, and the CLI warns after about 2 minutes without stopping the recording. That is useful for apps with longer loading, AI generation, or processing steps because ClipStitchr can cut pauses and waiting time during Quick Edit.
If the recording browser is not installed yet, interactive recording commands
ask to install it before recording starts. clipstitchr demo agent is
non-interactive, so it tells you the install command to run instead.
Creation commands let you start Stitchr and Swipr draft work from Terminal.
clipstitchr stitchr new creates today's Stitchr batch from recent UGC and
Demo clips. clipstitchr swipr new queues Swipr drafts using your dashboard
batch settings. Existing scripts can keep using the hidden stitchr batch and
swipr batch aliases. Queue commands add ready active content to your Post
Bridge queue without asking for a date or time. clipstitchr queue list shows
queued Stitches and Swipes coming up in the next 24 hours.
clipstitchr queue stitch picks the latest ready active Stitch unless you pass an ID,
clipstitchr queue swipe picks the latest ready active Swipe, and --all
queues active items one at a time. Dashboard browsing remains the best way to
inspect your whole Library. CLI Swipe queueing uses the saved rendered Swipe
image; open the dashboard when you need to render and queue the full carousel
or video version.
The interactive shell keeps one mounted terminal workspace with the current
menu, local product/repo/account context, a slash-command composer, suggestions,
action results, and keyboard hints. Existing guided questions temporarily take
input while an action runs. The menu collapses to one working line during that
handoff, keeps the question alive until it is answered, then opens the completed
result in the same workspace. When the terminal is too narrow, is not
interactive, or uses
--plain or NO_COLOR=1, the CLI falls back to simple prompt output. Direct
commands stay plain and script-friendly.
Setup detects common nested app folders like web/, infers the start command,
skips the product picker when your account only has one product, and prefers a
localhost URL that is already running.
Run clipstitchr products when you want a focused menu for listing products,
creating a new product, or choosing the product this repo should use. Direct
commands such as clipstitchr products list, clipstitchr products create,
and clipstitchr products use still work for scripts.
clipstitchr link connects the current repo to a product. clipstitchr init
does the same thing for developers who expect an init command. clipstitchr
unlink removes the repo connection without logging the whole machine out.
For native demos, browser recording works everywhere the recorder browser runs.
Manual native recording can capture an already-running iOS Simulator or Android
device/emulator; Android manual recording uses adb screenrecord when Android
tools are installed. Automatic OpenAI demos can use --surface macos-window on
macOS to select and control a visible window such as iOS Simulator, iPhone
Mirroring, an Android emulator window, or a selected desktop app. Run
clipstitchr native init once per Mac to install the helper under
~/Library/Application Support/ClipStitchr/, and run clipstitchr native init
--force to repair it. clipstitchr native check verifies the helper plus
Screen Recording and Accessibility permissions. On Windows, native
visible-window automation is not available yet, but browser demos still work.
Future adapter notes use windows-window for Windows visible-window control and
android-adb for direct Android ADB AI control; those adapters are not shipped
today. Window runs currently save screenshots/action logs; full helper-owned
MP4 capture is still handled by the manual native recorder path.
For local development against a preview app:
CLIPSTITCHR_API_URL=http://localhost:3000 npm run devThe CLI stores project settings in .clipstitchr.yml, saved walkthroughs in
.clipstitchr/demo-guides, local browser state in .clipstitchr, and machine
credentials in ~/.clipstitchr/credentials.json.
