Start with the thing you actually have.
A tutorial URL, phone MOV, or rough voice note can enter through Watch Source or Personal Source.
Native Mac source engine
Watchless
Turn what you watch and what you say into work your AI agent can build from.
Bring a tutorial, course module, Loom, call, or phone recording. Watchless keeps the source, writes the transcript, and returns the builder brief, script prompt, and next move in ordinary local folders.
Actual Watchless app
Native Mac productWatchless.app + DMG installer
Local workspace~/Documents/Watchless/
Real outputInspect the Printify sample
Default engineCodex CLI · no API key
What you can verify before you buy.
The page includes the product shape, the local folder shape, and a sanitized builder brief from a real capture. You do not have to take the promise on faith.
01Watchless.app
A packaged Mac app
Watchless ships as a native AppKit .app inside a DMG. It creates the customer workspace on first launch.
02See the folder
An inspectable source package
Media, audio, transcript, chapter clips, route files, and handoffs stay together in ordinary local folders.
03Open the sample
A real builder brief
The Printify sample keeps the generated structure and useful logic while removing private repo names, paths, and routing details.
04Your provider
A replaceable AI engine
Codex CLI is the local default. OpenAI API, Claude, OpenRouter, and Ollama remain optional provider lanes.
Two lanes: Watch Source and Personal Source.
Watchless is built for the desktop because source work means real files, transcripts, local folders, and agent-ready packages. The page works on a phone, but the product is for builders working at a Mac.
Watch Source
Turn outside material into a build package
Tutorials, talks, interviews, courses, demos, podcasts, internal training, and permitted research become chapters, transcript, notes, builder brief, project matches, and build prompt.
Personal Source
Turn your own voice into scripts and next moves
Phone MOVs, build diaries, founder notes, rough ideas, decision logs, and story capture become a transcript, personal-source handoff, and script prompt.
Local first
Keep the source package inspectable
Media, audio, transcripts, source notes, and knowledge files stay together so you can search them, hand them to an agent, or move them into your next project.
A beginning, a working middle, and a handoff.
These are real Watchless surfaces. Nothing is blurred into mood or cropped into decoration.
The source becomes a working object.
Media, transcript, routing, and next actions stay together in the local library.
Open the files. Build the next thing.
The builder brief gives your agent the logic, concepts, project fit, actions, and acceptance criteria.
Start with the thing you actually have.
A tutorial URL. A phone MOV. A rough voice note after a build. Choose Watch Source or Personal Source without changing how you capture.
- URL or local file
- MOV and MP4
- Your words stay attached
How it works
The app separates source capture from source use, so the file does not stop at a transcript.
Step 1
Choose the source lane
Paste an outside source or drag in your own MOV/MP4 recording.
Step 2
It transcribes + routes it
Splits chapters when useful, extracts the logic, and routes the source against the projects you've registered.
Step 3
You get the handoff
A builder brief, build prompt, script starter, or archive verdict your agent can use right away.
The more projects you register, the more it knows your stack. That's the part that doesn't leave with you.
The package keeps the source and the next move together.
One run hands you the media, transcript, and chapters when they help, plus source handoffs that turn the material into work you can act on.
your-source/
video.mp4 Mac-safe source copy
audio.mp3 clean audio
transcript.txt searchable text
personal-source.md first-person handoff when it is yours
script-prompt.md starter for posts, videos, or launch copy
knowledge/ the build breakdown for Watch Source runs
You keep the raw media, audio, transcript, and the files that make it useful.
Watch Source creates the full knowledge/ breakdown. Personal Source keeps your phrasing intact and turns the recording into a script-ready handoff.
Inside the knowledge folder
Nine files that take one watched source from "I saw that" to "here's my next move."
Understand itsummary.md
The gist — thesis, arguments, takeaways. Know what it's about in 30 seconds.
Understand itideas.md
The key ideas and concepts, pulled out and named.
Understand itactions.md
The concrete things the video actually tells you to do.
Matched to youproject-matches.md
Every project you've registered, scored 0–100 for how much this video helps it.
Matched to youbuilder-brief.md
A P0–P3 plan across your projects: what to change, which files, and why.
Matched to younext-actions.md
A roadmap ranked by impact vs effort — and the one move to make first.
Ready to buildbuild-prompt.md
A spec your AI agent runs cold: schema, files to touch, acceptance criteria.
It learnsoutcomes.md
Tracks what you actually did with it, so your library learns what pays off.
It learnsmetadata.json
Topics, frameworks, and creator tags that make your whole library searchable.
What stays on your Mac, and what does not.
Watchless is local-first. That means the files and library live on your Mac. A cloud AI provider still receives the text needed for analysis when you choose one.
Stays local
Media, transcripts, packages, and library
The working files remain in your local folders. You can inspect, move, archive, or delete them without a Watchless account.
Cloud engine
Analysis text goes to your selected provider
Codex, OpenAI, Claude, or OpenRouter receives the prompt and source text required to build the knowledge files. Their account rules apply.
Fully local option
Ollama keeps analysis on the Mac
Choose a supported local model when the source should not leave the machine. Capture and transcription also work without an AI breakdown.
Any useful source. Even the ones with no captions.
Most "summarize YouTube" tools break the moment a video isn't on YouTube or has no transcript. Watchless reads the audio itself, including your own local recordings.
Public links and permitted sources
Paste a supported link and keep the source, transcript, and useful chapters together.
A course you bought
No captions? It transcribes the audio and breaks it down anyway.
A Loom, a call, your own footage
Drag in the file and get a build plan, script starter, or personal source handoff.
Inspect the files, then read the brief.
This proof uses a real Printify capture package. The folder names and media sizes stay visible. The brief keeps the generated structure and removes private system details.
Input
"NEW Claude AI Affiliate Marketing Side Hustle…"
2:14:00 video
Output · after processing
P0 · Add Amazon Best-Sellers scoring to Product Scout
relevance 92 · with the files to touch + acceptance criteria
build-prompt your agent runs coldvideo.mp4 the source
audio.mp3 MP3 audio
transcript.srt captions
chapters/001-Introduction.mp4 18 MB
chapters/002-What_is_Personalization.mp4 28 MB
chapters/003-How_to_Set_Up.mp4 33.9 MB
knowledge/builder-brief.md the brief — sample on the right
knowledge/builder-brief.md · sanitized real output
Real output format, scrubbed for public viewing. Open the full sample
Send one source. Get the full build package back within 24 hours.
Nothing to install, no key to set up. Send a link or a personal recording you keep meaning to act on. It runs through the full engine, and you get back the complete folder with transcript, source handoff, and the right next file for the lane: builder brief, build prompt, or script prompt.
You send
One source, up to two, plus one line about what you're building.
A course module, a tutorial, a talk, a Loom, a call, or your own phone recording. Anything with signal in it.
You get back · within 24 hours
The full package: media, transcript, lane-specific handoff, project match where useful, builder brief, build prompt, or script prompt.
$99 · credited toward the full system when you upgrade. If the package isn't useful, the $99 comes back and you keep the files.
Send your source
first reply includes the invoice · you pay when it's queued
Pricing
It's a system, not a utility. One captured course can hand you an SOP, and one personal recording can become scripts, launch copy, or agent context. The price matches the saved hours and clearer inputs, not the tokens.
Start here
Proof Run
$99 · credited
- 1–2 of your sources, done for you
- Full package back within 24 hours
- Every dollar credits toward the Core System
- Not useful? Refunded — you keep the files
Core System
$897 / year
Codex CLI · no API key
- The Mac app and full capture workflow
- Watch Source and Personal Source modes
- Video, audio, transcript, SRT, chapter clips, and personal-source handoffs
- The nine-file knowledge breakdown, build prompt, and script prompt where available
- Codex CLI default, with API and local engines optional
- Source routing, searchable library, a year of updates
Custom Onboarding
$2,000+ · setup
- The engine tuned to your repos and projects
- Custom routing lanes and output rules
- First 10–20 captures tuned together
- 30-day calibration window
Codex by default, with provider choice built in.
The Core System uses your Codex login on the Mac, so the default path needs no API key. OpenAI API, Claude, OpenRouter, and Ollama remain available in AI Engine Settings. Provider limits and billing stay between you and that provider; Watchless adds no usage markup.