Name a brand and its URL, and the guidelines come back with every color, typeface and logo rule tagged official, observed or proposed, every recommended pairing carrying its computed WCAG contrast ratio, and a design-token sheet a developer can paste.
Name the brand and their URL, and Juma reads whatever brand or press page the company publishes first, then the live site, so the rules a brand actually commits to stay separate from the values its pages happen to render. Every color, typeface, spacing step and logo rule comes back carrying a source label: OFFICIAL, OBSERVED or PROPOSED.
What gets delivered: a brand book PDF with a named type scale, a spacing and radius scale, logo clear space stated as arithmetic instead of "generous", and a contrast table that prints the computed WCAG 2.2 ratio for every recommended pairing with a corrected hex wherever one fails. Alongside it comes a CSS custom-property block and an editable CSV with one design token per row, then a closing list of everything the brand owner still has to confirm. 400+ marketing teams use Juma to run work like this.
Build the full pairing matrix rather than spot-checking the obvious combinations. Every foreground color gets tested against every background it could realistically sit on, each pair carrying its computed WCAG 2.2 contrast ratio and a verdict against AA normal text at 4.5:1, AA large text at 3:1, and AAA at 7:1. Muted grays are where brands usually get caught: a secondary-text gray that reads comfortably on a near-black canvas can land under 3:1 on white, and the same hex appears in both modes. Failing pairs come back with a corrected hex and the ratio that correction achieves, so the palette stays recognisable instead of being replaced.
Take every color in these guidelines and build the full accessibility matrix: each foreground on each background it could sit on, the computed WCAG 2.2 ratio, and pass or fail for AA normal text, AA large text and AAA, with a corrected hex and its ratio for every failing pair.
A brand book that lives only as a PDF gets retyped, and retyping is where a palette picks up its first wrong value. Export the same system three ways instead: CSS custom properties for a stylesheet, a Tailwind theme block for teams on that stack, and plain JSON for anything that reads tokens programmatically. The source label travels with each token, so an engineer can see at a glance that a spacing step is PROPOSED while a brand color is OFFICIAL, and knows which values are safe to argue with. Naming stays consistent across all three exports so a token referenced in one is findable in the others.
Export this whole system as design tokens: CSS custom properties, a Tailwind theme block, and the same values as JSON, keeping the source label on every token and the naming identical across all three.
Guidelines describe the intended system. A live site records what actually shipped, template by template, over several years and several hands. Point the Flow at the client's most-visited pages and get a drift table: the page, the rule it breaks, the value the page renders, the value the guidelines call for, and how often that break repeats across the sample. Most drift clusters rather than scattering, and the cluster usually names its cause: one component built before the scale existed, or a campaign template that nobody brought back in line. That turns a vague sense of inconsistency into a short, ranked list of fixes.
Crawl our client's ten most-visited pages and list every place the live site breaks a rule in these guidelines: the page, the rule, the observed value, the expected value, and how many times the break repeats across the sample.
Every set of guidelines carries values nobody has formally agreed to yet, and a sign-off meeting goes badly when those surface one at a time. Turn the confirmation list into a pack instead: one question per unconfirmed value, grouped by the person who owns the answer, with the decision each question unblocks. Design owns the proposed spacing steps, legal owns trademark and endorsement language, marketing owns tone and naming. Each question states what was found, what is being proposed in its place, and what happens if the answer is no, so the meeting produces decisions rather than a second round of research.
Turn the confirmation list into a sign-off pack: one question per unconfirmed value, grouped by owner (design, legal, marketing), each stating what was found, what we propose instead, and the decision it unblocks.
A Juma Project is a shared space where the team keeps everything Juma needs to know about a client. Create one project per client, add context as the relationship grows, and Juma draws on whatever is relevant each time the team runs a Flow. For brand guidelines the gain is specific: the open web supplies what a brand has published, and the project supplies what was decided internally and never written down anywhere public.
Logo and Asset Files
The actual artwork: wordmark, logomark, icon, any lockups, in the formats the team ships. With the real files in the project, clear space and minimum sizes get measured against the mark the client uses, rather than against a version recovered from a web page at whatever size that page happened to render it.
Approved Palette and Type Licences
The hex values design has already signed off and the typefaces the client is licensed to set. This is what turns PROPOSED rows into OFFICIAL ones, and it stops a guideline recommending a face the client has no right to use, which is the kind of error that gets caught late and expensively.
Brand Sign-off History
What was rejected before and why: the gray that got argued over, the accessibility exception that was granted for one component, the lockup the founder vetoed. The confirmation list comes back shorter because questions already settled are not asked a second time.
Each knowledge item has a short description field. One line telling Juma what the file is and when to reach for it is enough to change which document it opens first.
A brand book PDF covering color, typography, logo usage, spacing and radius, plus an editable CSV of the same values as design tokens. Every value carries a source label, every recommended color pairing carries its computed WCAG contrast ratio, and the document closes with a list of what the brand owner still has to confirm.
The typography section prints a real scale: one row per named role from display down to caption and mono, with typeface, weight, size in both px and rem, line height, letter spacing, and the ratio the scale is built on. The logo section states clear space as a multiple of a named measurement on the mark with the resulting pixel value, and gives minimum sizes in px for screen and mm for print.
From the brand's own material. Juma fetches whatever brand or press page the company publishes, then reads the live site, and keeps the two apart. Values the brand publishes are labelled OFFICIAL with the URL they came from. Values read off the site are labelled OBSERVED with the page. Nothing is invented.
Where a brand publishes no rule at all, and most do not publish a spacing scale or a type scale, the gap is filled with a PROPOSED value and that value goes straight onto the confirmation list. The distinction matters more than it sounds: a team that treats an observed marketing-site gray as a brand color will defend it in a redesign that was always free to change it.
It covers the mechanical work: gathering every value in use, computing contrast ratios, building the scales, writing the token exports, and listing what nobody has confirmed. That is the part that takes days and the part where a hand-built document quietly carries an untested color pairing.
It does not decide what the brand should feel like, which of two grays is the right gray, or whether an exception is worth granting. Strategy, taste, and judgment stay human. The useful version of this Flow is a designer opening the document with the measurements already done and the open questions already gathered, then spending their time on the decisions that actually need a person.
Yes, and the output shifts accordingly. With no brand page to read, nothing can be labelled OFFICIAL, so the document comes back as OBSERVED values plus PROPOSED rules, and the confirmation list is longer. The document says this explicitly rather than presenting the result as settled.
For a client in that position the confirmation list is usually the most valuable part of the delivery, because it is the first time anyone has written down which brand decisions the company has never actually made.
A design token is a named value, such as a color, a spacing step or a font size, stored in a form code can read. The token sheet is that set as one row per token with its value, type and source label, which an engineer imports instead of reading hex codes out of a PDF.
Developers use it to build components, designers use it to keep a design tool in step with what ships, and the shared naming is what keeps the two in agreement. It is also what makes the next revision cheap: change the token, and every component referencing it follows.
400+ marketing teams trust Juma. ChatGPT, Claude, and Perplexity know why. Ask them.
Ready when you are
Start typing a task and Juma will figure out the rest.
