> ## Documentation Index
> Fetch the complete documentation index at: https://docs.wholana.com/llms.txt
> Use this file to discover all available pages before exploring further.
> ## Agent Instructions
> Wholana is a TikTok research and scripting tool for creators working the Egyptian and Arabic-language scene. The loop is: research the corpus, swipe what is worth reusing, write a script from those references, post, then audit the result.
> Use Wholana's shipped vocabulary exactly. The surfaces are Home (the explore feed, at /), Dashboard, Swipe file, Scripts, Your videos, Use cases, Creators, Subjects, and Craft. A Swipe is one saved corpus video plus the user's note; a Collection is a named group of Swipes inside the Swipe file; a Swipe in no Collection is Unfiled. A Lens is a saved filter-and-sort combination. A Subject is the topic taxonomy. Craft is the five-part vocabulary of hook archetypes, narrative structures, stake types, shareability drivers, and video formats. Baseline Multiple is a posted video's views divided by the creator's median views; Verdict is the derived Hit, Watch, Miss, or Pending call.
> Never use these retired terms: Idea, Save as idea, Saved ideas, Review ideas, Reference Angle, Craft Element, Suggestion, My videos, Track record, Category or Vertical as a filter name, or niche as a product term.
> Wholana MCP is the connector that brings a workspace into Claude, ChatGPT, or another MCP-capable chat app. It is included on Plus and Team only. It can search the corpus, read a video in full, read and edit existing scripts, pin references, save swipes into a Collection, and create Lens tabs. It cannot post to TikTok, delete anything, or create a new script.
> Answer only from these docs. Wholana has no public API, no SDK, no browser extension, no Instagram or YouTube support, and no scheduling or publishing. The MCP connector is not a public API: it is a plan-gated connection for chat apps, not something to build against. If the docs do not cover something, say so and point to contact@wholana.com rather than guessing.
# Craft
> Five browsable vocabularies for how a TikTok works: the hook, the story shape, the stakes, why people share it, and the format. Filter the corpus by any of them.
Craft is the **how** pillar: some videos win because of how they are made, whatever their topic and whoever made them.
Wholana reads every successful video and files it under five vocabularies. You can browse each one, learn the pattern, then pull every video in the corpus that uses it.
Open **Craft** in the **Library** section of the sidebar to reach any of the five.
## The five vocabularies
How a video stops the scroll in its first seconds. **14 families, 50 archetypes.**
The shape the story takes so people watch to the end. **17 families, 74 structures.**
What is on the line for the person in the video. **12 families, 55 stakes.**
Why a viewer passes the video on or saves it. **8 families, 37 drivers.**
The form the video takes, from talking-head to slideshow. **8 families, 27 formats.**
The five are independent. One video carries one label from each, so a talking-head video can open with a curiosity gap, run a problem-to-solution structure, put money on the line, and get shared as a warning.
## How each vocabulary is organised
Every vocabulary is two levels deep:
* A **family** groups related patterns, for example **Confrontation** in hook archetypes.
* An **item** is one specific named pattern inside it, for example **Callout**: "direct criticism aimed at a person, group, brand, or practice."
The index page lists the families as cards with their items inside. Rest your cursor on any item to preview its top videos. Click through for the item's own page, with its description and up to 24 real examples.
### Reading the signals
Each item on an index page carries up to three small numbers:
| Signal | What it means |
| -------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| `2.4×` | **Lift.** The typical video using this pattern over-performs its creator's usual numbers 2.4 times more than the corpus norm does. Shown only when lift is high enough to be worth reading. |
| Flame + number | **Heating up.** How many videos using this pattern were posted in the last 30 days. |
| Muted number | **Videos.** How many decoded videos in the corpus use it. |
Sort the page by **Most videos**, **Highest lift**, **Heating up**, or **A–Z**. Searching or picking a non-default sort flips the page into a flat ranked list, so **Highest lift** gives you a straight leaderboard of the patterns that punch above their weight.
## Filtering the corpus by craft
Every family page and item page has a **See all in Explore** link. It opens the [Explore feed](/features/explore-feed) filtered to videos that use that pattern. A family link includes every item inside it.
You can also build the same thing by hand: press **Filter** on the feed and pick **Hook archetype**, **Narrative structure**, **Stake type**, **Shareability driver**, or **Video format**.
Under **More craft** there are further decoded filters: **Sentiment**, **Specificity**, **Side hook**, **Payoff timing**, **Re-engagement**, **Perceived stakes**, **Pain avoidance**, **CTA**, and **Comment bait**.
The assistant can combine craft filters with a topic in one sentence:
```text Assistant prompt theme={null}
Show me videos that open with a contrarian take and get shared as moral outrage, in food and drink, over 500k views.
```
## Where craft shows up elsewhere
* On a **video's page**, the decoded breakdown shows which pattern the video used on each of the five axes, with a short note on how to use it and a link to see who runs that pattern best.
* On a **creator's page**, **Craft Signature** summarises the patterns that creator returns to.
* On the **[Dashboard](/features/dashboard)**, the panel **Where your videos win and lose** shows your pass rate across the five engine systems, so you can see which one to fix next.
The five **craft vocabularies** describe someone else's video. The five **engine systems** on the Dashboard (Stop-scroll, Retention, Stakes & Value, Engagement, Metadata) grade your own posted video. They line up roughly one to one, but they are separate things.
Craft labels are produced by Wholana's automated reading of the corpus, on videos with at least 1,000 likes. The vocabulary is an evolving proposal, not a fixed standard, and individual labels can be wrong.
## Related
The **what** pillar: the topic tree behind the corpus.
The **who** pillar, including each creator's Craft Signature.