The product

Brand guidelines

The Guidelines library is your workspace's brand memory: the tone, colors, facts, and writing rules your agent must read before it drafts a single word. This page is the reference: what the library holds and how it binds your agent. For the hands-on setup, follow the guide.

What the library is

Every workspace has one Guidelines library: brand and context memory for that one brand, curated by you. Boards opt in to it. On a board with guidelines enabled, your agent is required to read the library before writing anything on that board, and to treat what it finds as binding rules rather than reference material it might consult when it feels like it.

That one design choice is why the tenth draft sounds like your brand instead of drifting back to generic AI copy. You state the rule once, at the workspace level, and it applies to every email, caption, and ad on every opted-in board, without you repeating it in every chat.

Guidelines sit alongside two other ways of steering your agent: templates carry how a kind of campaign is run, and card comments steer one piece of work at a time. Guidelines are the layer that doesn't change per campaign: who your brand is, wherever your agent is working.

i

Reference vs. tutorial. This page explains the library. Set guidelines for a workspace is the doing page, with copy-paste prompts that let your agent research your brand and build the blocks for you.

The five block kinds

The library is a stack of small blocks, each one of five kinds:

Text
Written rules and facts: tone of voice, do/don't lists, what you sell and to whom. Most libraries are mostly text blocks.
File
A real artifact to use or imitate: your logo, your fonts, an example email in your house style. Files are stored privately inside your workspace.
Link
A public source of truth, such as your brand page or a published style guide, that your agent can go read.
Color palette
Named colors with exact hex values. "Our red" stops being approximately red.
Voice
The "Prevent AI language" preset: a set of switches that stop copy from reading like it was machine-written. Detailed below.

The voice block

One block kind earns its own section. The voice block is a preset with five switches, and every switch you turn on becomes a concrete writing rule your agent follows verbatim:

  • No AI punctuation: ends the em-dash-in-every-sentence habit.
  • Cut fluff: strips empty filler and throat-clearing openers.
  • No AI hype: bans the "unleash / elevate / game-changing" register.
  • No formulaic structure: blocks the "It's not X, it's Y" tic and lists-of-three padding.
  • No emoji: for brands that don't speak in rockets.

When a stakeholder asks "won't it sound like ChatGPT?", you can point at this block: a switch you flip, enforced on every draft in the workspace, instead of a hope that the model behaves.

Per-board control

Guidelines apply per board. When you create a board, a checkbox lets you opt that board out, which is useful for a scratch board where you're just thinking out loud, or a competitor-research board where writing in your own voice would be beside the point. On every board that stays opted in, your agent reads the library before it writes, every session, without being reminded. Your agent can also see whether a board has guidelines enabled, so there's never ambiguity about which rules are in force where.

Who maintains them

You do, on the Guidelines page in the app. It opens with a suggestion gallery of eight starters: Prevent AI language, brand logo, tone of voice, example HTML file, color palette, brand fonts, do/don't rules, and a public source link. A useful library is a few clicks away instead of a blank page.

Your agent can help: ask it and it will create new blocks and update existing ones, for example after it has analyzed your website, or when a test draft shows a rule is missing. The workspace guidelines guide walks through that agent-driven setup end to end. Either way, the library stays yours: small, current, and only as strict as you want your brand to be. Blocks are editable at any time. Because your agent re-reads the library every session, a change takes effect from the next thing it writes.