Brand
Brand (/brand) holds what every draft and every revision reads. It is why a draft sounds like your team rather than a model. Viewfield fills it from your website during the welcome, and from then on you change it two ways that write to the same place: edit it here, which applies at once, or tell Ask Viewfield in plain words ("never say seamless", "always open with the customer's problem", "warmer, less corporate") and confirm the card it shows. Either way the next draft obeys; there is nothing else to apply.
Everyone in the workspace can read every section. Managers change them; a contributor sees the same pages without the controls.
The board
Brand opens on a board under "How you sound", in this order:
| Tile | What it holds | Opens |
|---|---|---|
| Your voice | Your voice as its blob, and the four dials | Voice, Tone |
| Never say | The words and phrases every draft avoids | |
| Always | Your standing rules | Writing rules, Checks |
| Sounds like | The few words your voice sounds like, the sample line (rewritten when your brand changes, or on request), and the tone in words, which you edit in place | |
| Learn from | Pieces that sound exactly like you | Pin a piece (managers) or See examples |
| Kinds | A rule per kind of draft | Platform rules, Add a custom kind |
| Basics | Your tagline, description and values | Description, Beliefs and stance |
| Docs | Product sheets, proof points and anything your site does not say | All docs |
The voice
The voice is four dials and the words that go with them:
| Dial | From | To |
|---|---|---|
| Formal to casual | Formal | Casual |
| Plain to bold | Plain | Bold |
| Serious to playful | Serious | Playful |
| Technical to simple | Technical | Simple |
Each dial reads out as words ("Mostly casual"). The writer reads each dial as one plain instruction for the band it sits in (five bands, from one end to the other, the middle one included), so a nudge inside a band changes nothing and a draft is never different for a reason you cannot see. A dial you have never set adds nothing.
Releasing a dial saves it at once. The Sounds like sample line follows your whole brand: release a dial, or save a new description, tagline, tone or name (here, on Description, or by telling Ask Viewfield), and the line is rewritten from the new words a moment later with one model call on your allowance, with "Rewriting…" beside it while it is; several quick edits make one rewrite. A dial on its own keeps what the line is about and changes how it sounds; a new description writes it fresh. If a rewrite fails, the old line stays, dimmed, with Try again, and the refresh button beside Sounds like writes a new sample whenever you want one. The line never uses a word on your never-say list.
The longer voice description lives on two pages: Voice (your professional background, personality traits, expertise, communication style, perspective and relationship with your audience) and Tone of Voice (voice characteristics, the tone spectrum, language preferences, emotional qualities, formality, the elements unique to your voice, and practical guidance). Each keeps its version history.
Writing rules
Writing rules (/brand/rules) holds the hard constraints: sections for grammar and punctuation, formatting, voice and perspective, sentence structure, word choice and content structure, then:
- Banned phrases (the board's Never say): the words and phrases every draft avoids, up to 200, each at most 160 characters. The never-say marks in the editor catch any that slip through. Add one, or remove one to allow it again.
- Standing rules (the board's Always): "always sign social posts with the team name", "lead with the problem, not the feature", each at most 400 characters. Add, edit in place, delete.
Ask Viewfield writes here too, on a card you confirm: "never call it a platform" joins Never say; "stop opening with a question" joins the rules; "allow the word platform again" removes one. Agents do the same over MCP with set_writing_directives.
Kinds
Kinds sets how each kind of draft is shaped: Blog post, LinkedIn post, X post, X thread and Newsletter, plus the custom kinds your team adds. A rule, on its Platform rules page, has a Tone, Format rules, Openings and a Typical length (a hint to the writer, never a limit). Viewfield's defaults apply until you change one.
A custom kind (Add a custom kind) is a kind of your own with its own rule: release notes, an investor update, podcast notes. A workspace can have up to 20. A custom draft names its kind and follows that rule. A custom kind a draft was written in during the last 30 days cannot be deleted.
Checks
The Checks run over every draft and every revision after it is written (the Skills page, /brand/skills, under "Checks on every draft"). The built-in checks are Viewfield's:
| Check | Default | What it does |
|---|---|---|
| Anti-slop | On | Strips the tells of AI writing (em dashes, filler vocabulary, "it's not just X, it's Y", throat-clearing, restated conclusions) and enforces specific, natural prose, with a lint pass and one targeted repair after every draft |
| Casual language | Off | Keeps the register conversational: contractions, short words, direct address, no corporate speak |
| Technical accuracy | Off | Grounds every factual claim: no invented numbers, quotes, sources or capabilities |
| Answer engine readiness | Off | Structures long-form content so AI answer engines can cite it: conclusion-first sections, question-shaped headings, self-contained passages, citable specifics |
Your voice always outranks a check: a word your lexicon loves is never flagged. A manager can switch the checks off all at once. A check of your own (a SKILL.md-style instruction) is added by asking Ask Viewfield or with add_skill over MCP, and then switched or deleted on this page like the others.
Learned from your edits lists the rules the checks picked up. When you mark a draft Final, what you kept and what you rewrote teaches candidate rules; recurring asks in the chat ("shorter hooks") are distilled into rules; a rule the lint keeps catching weighs more. A candidate waits for a second edit or your confirm. Each rule offers Confirm (apply it to every draft), Mute (stop applying it; a muted rule never comes back on its own), Apply again and Delete. Its weight is changed through Ask Viewfield or set_memory_weight over MCP.
Over MCP: list_skills, set_skill_enabled, add_skill, update_skill, delete_skill, list_memories, set_memory_weight, mute_memory, remember_rule, and guard_review, which runs the checks over any text or public page.
Basics: description, perspective and stance
Description (/brand/description) is the brand's name, website, tagline and description, as every draft reads them. Changing the website can read the site again; it never replaces anything a person accepted or wrote.
Perspective and stance (/brand/perspective, the board's Beliefs and stance) holds your core beliefs, mission, vision, values, unique perspective, brand personality and your positioning statement: what you argue from. A draft reads it; nothing else in the product changes it on its own.
Docs
Docs is the library a draft can quote: product facts, proof points, messaging frameworks, past posts. Add a doc three ways in the app (Upload a file: PDF, Word, text, Markdown or Excel, up to 15 MB; Paste text; From a link, which reads one public page), or over MCP with upload_document.
Turn on Always read for the docs every draft should see. A draft reads up to 8,000 characters of each always-read doc and 24,000 across all of them, and the page shows how much of that budget is used. The other docs are found by search: Ask Viewfield and agents search the library by meaning.
Viewfield already reads your site, so a doc is for what the site does not say.
Learn from
Learn from holds past pieces pinned as how the brand sounds: "Pieces that sound exactly like you. Every draft reads them, and never copies them." Pin up to five; the opening of each, as it read when pinned, is kept, and the newest three ride into every draft as something to sound like and never to copy. Unpinning removes the example and leaves the piece.
The companies AI answers name
The list of companies AI answers may name instead of you is not on Brand: it is the Who else gets named tile on the Visibility board. Each company there has its own page under Brand (/brand/competitors/<id>), with its website and its class (Direct rival, New entrant, Adjacent, Substitute, or Not a rival, so AI visibility stops suggesting it). Nothing is tracked, researched or watched about them.
Versions
Voice, Tone of Voice and Writing rules keep version history, and so does every item. Over MCP: list_engine_versions, save_engine_version, restore_engine_version, rename_engine_version, delete_engine_version, list_item_versions, restore_item_version.
Over MCP
get_engine_config reads every section. update_voice, update_brand, update_positioning and update_writing_rules merge the fields you provide, so a partial update never wipes a value; set_writing_directives teaches a ban, a rule or a tone; set_platform_tone sets a kind's rule; add_competitor, update_engine_item and delete_engine_item manage the company list and the docs; ai_update_engine_item rewrites an item from an instruction. Every write here is manager-level, from the app, the MCP server and Ask alike. See MCP overview.