What can Whalory help you write?

Install Whalory as a skill in your AI assistant. It guides the assistant to clarify your brief, draft and revise, follow your voice, and review the result.

From the first question to the final review

Whalory gives your AI assistant a writing process. Your assistant’s AI generates the text; Whalory’s local checks run without an API key.

Start with the right information

Set the audience, purpose, and facts the text needs. Whalory guides your assistant to ask useful questions before drafting and mark information that is still missing.

Prepare a brief

Draft something new or revise what you have

Turn your information into a draft, or make an existing text shorter, better organized, and suited to its reader. Keep the meaning you intended.

See the writing process

Find language issues

Local checks flag issues such as stock phrases, text length, and Persian character and spacing errors. They help you spot problems; factual accuracy needs a separate review.

Try the text checks

Keep your voice consistent

Define how formal or friendly the writing should be, which words fit, and which phrases to avoid. Give these preferences to your assistant for use in future drafts.

Build a voice profile

Adapt one message for different places

Keep the facts while adapting the delivery for an email, a text message, a website, or a social post. Account for each format’s length, structure, and call to action.

Compare writing formats

Give the draft another read

Check the draft against the original brief, supplied facts, and chosen voice. Whalory provides a review process; a separate editor agent depends on what your assistant supports.

Compare two revisions

What can you do in the website workspace?

Prepare a brief, build a voice profile, and export the details for your assistant. Live AI text generation is currently disabled in the website workspace.

Open the writing workspace Browse writing formats

The interactive examples below are prepared texts. Compare their formats and try language checks that run in your browser.

One brief for your product launch

  1. Supply the details

    Give the product features, price, audience, voice, and intended channels.

  2. Prepare the set with Pro

    The campaign command guides your assistant through planning, drafting, and reviewing website copy, a social caption, and a text message.

  3. Check before publishing

    The comparison tool flags changes to numbers and names. You still check the meaning and verify the facts.

This is an illustrative workflow. Your assistant generates the copy; quality depends on your information and the model.

Prepare your own campaign brief

For recurring client or brand work, Pro adds a campaign command, version comparison, and batch checks. Core supports general writing and rewriting in your existing assistant.

Compare the editions in detail

Channel lab

One brief. Three places to publish.

A text message is read for the news; a product page is read for the details. The same facts need a different draft for each place.

Business

What the brief says

All three drafts use these facts and nothing else.

Name
Linden Coffee
News
Opening at 7 a.m. instead of 8 a.m.
Why
A few regulars asked for coffee before work
Coffee to go
From the side window
Start date
not supplied
Fill in the missing detail

Only this value goes into the draft, exactly as you type it. Leave it empty and the gap stays in brackets.

Destination

Prepared exampleCaféInstagram caption

Your accountThe post’s photo is not part of the example.

Starting Placeholder: [start date] (note 3), Linden Coffee opens at 7 a.m., an hour earlier. (note 1)

A few of you asked for coffee before work. (note 2) Now there’s time for one.

In a hurry? Grab your coffee to go from the side window.

198 characters, 0 hashtags

Still open: [start date] This draft still has gaps in brackets. Copy or download it as a draft, or fill them in first.

Clears your value and brings the placeholder back.

Why it reads this way

Prepared review notes
  1. Moved

    News in the first line

    The new opening time is the news, so it sits above the “more” cut. “An hour earlier” saves regulars from doing the math between 7 and 8.

  2. Kept

    The reason you gave

    The brief says a few regulars asked for this. Their request explains the change in their own terms and promises nothing new.

  3. Placeholder

    Start date kept open

    No start date was supplied. The caption keeps the gap in plain view, so a guessed date cannot go out as news.

Instagram caption

The feed cuts a caption off behind “more”, so the news goes in the first line. Hashtags go at the end, five at most.

Rules for this destination

Runs in your browser

These checks run in your browser on the text as it stands. They check form, not truth: whether a fact is right is still your call.

  • No gap left in brackets

    Look again

    Still open: [start date].

    Sources: whalory/SKILL.mdwhalory/scripts/lint_en.py

  • Nothing from the avoid list

    Pass

    None of the phrases on the avoid list.

  • Consistent quotes, apostrophes, and spacing

    Pass

    One style of quotes and apostrophes, single spaces.

  • Caption of 2,200 characters at most

    Pass

    Within the limit: 198 of 2,200.

    Source: whalory/data/channels/social.json developers.facebook.com

  • Five hashtags at most

    Pass

    Within the limit: 0 of 5.

    Source: whalory/data/channels/social.json threads.com

Try a flawed version

Runs in your browser

Here is the same message, written badly on purpose. Edit it until every check passes.

The flawed version loads with the page’s script, because the checks run in your browser.

Capabilities

The craft behind the copy.

Eight working habits Whalory brings to the AI you use. We wrote each example for this page.

Core, the free edition, includes the first seven; the eighth is a study that has not run yet.

  • Clarify the brief

    Whalory asks only when the answer would change the text, and never more than three questions at a time.

    Illustrative example

    1. Brief

      Product page for our handmade candles. Mention the fast delivery.

    2. One question

      Do you know when an order arrives, or only when it leaves you?

      A customer buying a birthday gift may read “fast delivery” as a promise that it will arrive in time.

      Pick the shop’s answer

    3. Grounded sentence

      We ship every order within one business day.

      Arrival stays out of the sentence until someone confirms it.

  • Respect supplied facts

    A number, a name, or a promise missing from your facts gets a bracket, never a guess.

    Three rules that never change

    1. Nothing made up: numbers, quotes, reviews, and results come only from your facts.
    2. No machine patterns, such as straw-man contrasts or stacked adjectives.
    3. No claim without evidence the brand can show.

    Your facts

    • Handmade soy candles
    • Scent: cedar

    First draft

    Our cedar candles burn for 60 hours.

    After the check

    Our cedar candles burn for [confirm: burn time].

    Burn time is not in the facts, so “60 hours” was a guess.

  • Fit the publishing channel

    A long text goes out in several parts. This counter splits it by encoding, with the same rules as Whalory’s own counter.

    Counted
    116
    Messages
    1
    Encoding
    GSM-7, the basic alphabet

    One message holds 160; once it splits, each part holds 153. Room for 44 more before the next split.

    Messages: 1. Encoding: GSM-7, the basic alphabet.

    The numbers follow the GSM-7 and UCS-2 rules in Whalory’s counter. Your SMS provider may count or bill differently.

  • Preserve brand voice

    A voice profile keeps your dials and your words. Whalory reads it before every task.

    From Whalory’s own profile

    Its eight dials

    Warmth
    3 / 5
    Formality
    3 / 5
    Humor
    1 / 5
    Narrative
    3 / 5
    Sentence length
    3 / 5
    Loud marks
    0 / 3
    Jargon
    2 / 5
    Rhetoric
    2 / 5
    SayDon’t say
    voice profilebrand voice document
    Whalory ProWhalory Premium
    Build your own profile

  • Keep the details consistent

    The English checks catch mixed US and UK spelling, stray straight quotes, and a numeral that opens a sentence. Persian copy gets checks of its own, for half-spaces and Persian digits.

    Before

    12 "Cedar" candles are back. Today’s orders ship by 5pm & the rest ship tomorrow, each in a gray box with a personalised label.

    After

    Twelve “Cedar” candles are back. Today’s orders ship by 5pm and the rest ship tomorrow, each in a gray box with a personalized label.

    Rules that fired, by their linter names

    • en-numeral-starta numeral opens the sentence
    • en-quote-mixstraight and curly quotes in one text
    • en-double-spacetwo spaces after a full stop
    • en-ampersandan ampersand in running text
    • en-spelling-mixUS and UK spelling in one text

  • Write naturally in English

    Stock phrases that make English copy sound generated get flagged by name. Try a sentence of your own.

    Patterns found: 3

    • en-cliche-openWarningIn today’s fast-paced worldStock opening. Start with the fact.
    • en-buzzwordWarningunlockBuzzword. Say what it does, and for whom.
    • en-buzzwordWarningseamlessBuzzword. Say what it does, and for whom.

    Patterns found: 3

    Twelve rules from Whalory’s English linter run here, in your browser. Nothing leaves the page, and the full linter checks much more.

  • Check and review

    Scripts check what a machine can count. A careful reading checks the rest, and the note says which checks ran.

    Flagged line

    Pass your driving test in 30 days.

    Facts supplied: a 30-day driving course. No pass rate.

    The note names the quality check (QA) that ran

    • QA: script
    • QA: manual
    • QA: manual (script not run)

    Who reviews: the Claude Code plugin includes a separate editor agent in both Core and Pro. In other hosts, independent review depends on their agent capabilities; otherwise the assistant reviews its own draft in a separate pass.

  • Study pending

    Inspect evidence

    A comparison with raw AI, with its method fixed before any run.

    • Per language: 26 briefs, from low to high risk
    • Blind pairwise preference by human and AI judges
    • Ties and losses reported as they come

    No run has taken place, so there are no results to show.

    Read the method