# Spectrea Brand Guide > Composable intelligence platform — turns scattered information into compounding intelligence. This file is a generated ROUTER: it tells an AI tool which brand document to load, and in which order. Every document below derives from the same canonical data. **Source of truth:** brand.ts is the canonical brand data. The app renders it; the guide (brand-guide.md), llms.txt, the PDF, and generated assets are derived mirrors. On any conflict, brand.ts wins. Capability claims are anchored to the ratified product vision; every buyer-facing capability claim is checked against the product roadmap before it ships. Execution status is tracked there, never here — this file is the brand, not the release note. (v2.15.0, 2026-08-18) ## Load order (for AI tools) 1. [Brand contract (JSON)](/brand-contract.json): hard machine constraints — names, category noun, banned terms, palette hexes, design tokens, logo numerics, guardrails. 2. [Pre-flight checklist](/brand-checklist.md): classify your surface, then run its checks before generating anything. 3. [Few-shots](/brand-few-shots.md): contrastive on/off-brand pairs — match the on-brand register. 4. [Agent rules block](/brand-agent-rules.md): drop-in CLAUDE.md/AGENTS.md block for repos that produce Spectrea-branded output. 5. [Full brand guide (Markdown)](/brand-guide.md): the complete narrative system — foundation, voice, naming, logo, colour, gradients, typography, iconography, illustration, motion, components, communications, governance. 6. [Full brand guide (PDF)](/brand-guide.pdf): print/offline mirror of the same content. ## Hard key facts - **Name:** Spectrea (/spek-TREE-uh/) — never: "SpectreAI" (not the name) · "Spectre" / "Spectra" (different words — not the name) · "spectra" (missing the e — typo) · "Spectrea AI" as the lead name (AI is not the brand) · "SPECTREA" (all-caps styling) · "spectrea" lowercase in headings or prose (lowercase belongs to the logo wordmark treatment only) · "The Spectrea" (no article) · Random internal caps ("SpecTrea"). - **Category:** The category noun is exactly "Composable intelligence platform" (lowercase in running prose) — never a substitute: "composable knowledge platform", "knowledge platform", "knowledge management", "AI assistant", "automation tool". - **Tagline:** "We connect the dots." - **Promise:** Turns scattered information into compounding intelligence. - **Voice formula:** Plain words, real specifics, room to breathe. Tech earns its place by being checkable — shown once, where the reader looks for it. - **Attention rule:** Show the work once, where the reader looks for it. "Earn the jargon" governs what may enter the copy — not how often it repeats. One claim, one proof per surface: the mechanism lives one step from the claim (the next sentence, a subhead, an expandable) — never welded to every sentence. If two sentences in a row prove the same claim, cut one. The claim—dash—mechanism shape is a tool, not a rhythm. - **Never-use words:** AI-powered (overused), next-gen, state-of-the-art, world-class, breakthrough, unprecedented, game-changer, revolutionary, seamless, cutting-edge, synergy, leverage (as verb), agentic, AI-first, copilot (as generic noun), 10x, supercharge, effortless, magical. - **Density:** Max two privileged words per paragraph, one "living" / "alive" per page. If a paragraph survives losing every privileged word, the prose is doing its job — if it collapses without them, rewrite. - **Hero open:** Open on the outcome, written as a reveal — the moment of new sight. Never open with the mechanism, the pain alone, or a platitude. Ground it next with a specific the reader can check; close by equipping the next step. The gesture adapts to any surface — a problem may set up the reveal, proof may carry the grounding — but the open always lands on the outcome. Exception: graph-native jobs (exploration, visualisation), where the graph is the star and may lead. - **On-ramp:** Outcome-first hero per the canonical heroOpen rule: the hero leads with the entry job and its outcome, written as a reveal. One coined whole-product frame is sanctioned (ratified 2026-07-19): "the operating system for collective intelligence" — distinct from generic "AI OS", always paired with plain language that carries the meaning. Balance doctrine (ratified 2026-07-19): avoid-list terms are not blacklisted — one may label a segment-matched surface when a plain-language claim carries the meaning in place; never expect it to carry weight with unfamiliar audiences, and never lead a cross-segment hero with it. Hero example: "Drop in your docs — get answers that show their sources, and a system that gets sharper every week." - **Anti-values:** Never replaces human judgment · Never leaves anyone behind · Never complex or bloated · Never a black box. - **AI naming:** Three registers. (1) Homepage/hero narrative: "your AI" is sanctioned to match the hero register. (2) Product and trust surfaces: the AI is "the assistant" — lowercase, generic; as a first-class feature name it is "Assistant" (Title-Case). (3) Technical and architecture writing: "agent" is the umbrella term for an accountable AI actor (delegate or standing agent), per the vision taxonomy — the assistant is the delegate shape's product realization. Never "copilot" (generic or name), never "Spectrea AI" as a name. The assistant suggests, surfaces, and drafts — it never silently "handles it for you" or "takes over". Automations act within guardrails the user sets: the user decides what runs on its own and what checks with you first (ratified 2026-07-19, per the vision's governed-autonomy model). Never describe autonomy without naming its guardrail; never use approval-begging framings ("waiting for your yes"). Allowed verbs: suggests, surfaces, drafts. Forbidden verbs: handles it for you, takes over. - **Compounding claim guardrail:** Use "compounding intelligence" only with its defensible mechanism named — provenance, per-viewer access, and the closed loop that feeds outcomes back in. The phrase is contested in-market (noted 2026-07-03); the mechanism, not the slogan, carries the claim. - **Spectrum accents:** Cobalt #4271DF · Teal #00B6A0 · Amber #E19000 · Rose #F24260. - **Warm Blend neutrals:** Graphite #212226 (primary) · Canvas #FDFDFB (background) · Cloud #F4F4F1 (surface) · Ink #18181C (text) · Iron #46464B (body-emphasized) · Slate #6D6D72 (body-secondary) · Pewter #97979E (muted). - **Dark mode tokens:** bg #18181C · surface #212226 · text #F4F4F1 · muted #B0B0B6 · border #2E2F35. - **Gradient:** #4271DF → #00B6A0 → #E19000 at 135° — never for buttons, text colour, body backgrounds, borders, or small icons. - **Logo constraints:** exactly 10 dots (radius 3.9607, stroke 9.0531); 2 lockup forms; primary dots #A3A3A3; Circle when contained — never a squircle. Clear space: 0.5× the mark height on all sides. Watermark ≤ 20% opacity. Co-brand: Use the ink variants (#3 mark / #4 lockup). Match the partner mark by optical height — scale so perceived heights are equal, not bounding boxes — and align on the marks' vertical centres. - **Typefaces:** Albert Sans (headings) · Lexend (body) · JetBrains Mono (code). - **Accessibility floor:** WCAG 2.2 AA (contrast ≥ 4.5:1 normal / 3:1 large+UI). - **Personality:** Warm · Perceptive · Grounded · Adaptive (4 traits, each with a guardrail). ## Visual assets Standalone SVGs live in /brand-assets/ (regenerated by "npm run generate:assets"); core files: logo-mark-cool.svg, logo-lockup-gradient.svg, logo-lockup-ink.svg, logo-lockup-white.svg, swatches-spectrum.svg, swatches-neutrals.svg, type-scale.svg. Generating illustrations: use the DNA block and register sentences at [/illustration-prompt.md](/illustration-prompt.md) (generated from brand.ts — the same doctrine is in this contract under "illustration"). ## What is deliberately not here (internal tier) Fields listed here render only to internal/ artefacts. Adding a field is a ledger decision; generators must consult this registry, and a public artefact carrying an internal field is a build defect. Registered fields: trustCopy, brand.positioning.fullShapeClaim, brand.differentiatorGuardrail, brand.antiBrands, brand.audienceMechanics, ratificationLedger. The internal-tier fields named with this rule are absent from every public artefact by design, not by omission. If you are working from a vendored snapshot, do not reconstruct them from an older snapshot, a cached copy, or memory, and never author replacements freehand — trust, security, and compliance wording is counsel-gated. When a surface genuinely needs one, request the internal brand hand-off (the internal/ artefacts, delivered locally) instead of inventing it. ## Vendoring this into another repo Consumer repos (the Spectrea app and website) vendor a snapshot rather than importing canon. Take the set together — /brand-contract.json, /brand-checklist.md, /brand-few-shots.md, /brand-agent-rules.md, /llms.txt, /brand-guide.md — plus /brand-conformance.mjs, the generated checker that enforces this canon in a consumer CI. All of them are generated at the same canon version, so a partial re-vendor mixes versions. - **Current version:** 2.15.0 (2026-08-18), published at https://branding.spectrea.com. - **Pin to a release, not to a date.** Selected canon versions are cut as GitHub releases at https://github.com/Matchwise/spectrea-branding/releases — each one is a version consumers are meant to adopt, with the whole set attached as a single snapshot archive. Not every version bump gets a release; the live files above always reflect the newest canon, released or not. - **Staleness test:** `curl -s https://branding.spectrea.com/brand-contract.json` and compare `version` with the vendored copy's. - **Never hand-edit a vendored file.** It is generated; the next re-vendor silently drops the edit. Changes go to src/data/brand.ts in the Spectrea branding repo and come back through regeneration. - **The checker is versioned too:** its rules are baked in at generation time, so an old copy enforces old canon. Re-vendor it with the snapshot. ## Live interactive pages (SPA — JavaScript-rendered) NOTE: the routes below are client-rendered React routes; most AI crawlers cannot execute them. Use the Markdown/JSON documents in the load order above instead — they carry the same canonical content. Routes listed for humans and browser-capable agents: - [Overview](/#/) - [Foundation](/#/foundation) - [Brand Story](/#/foundation/story) - [Positioning](/#/foundation/positioning) - [Voice & Tone](/#/foundation/voice) - [Naming](/#/foundation/naming) - [Logo](/#/logo) - [The Logo](/#/logo/primary) - [Variations](/#/logo/variations) - [Guidelines](/#/logo/guidelines) - [Misuse](/#/logo/misuse) - [Specs & Export](/#/logo/exploration) - [Color](/#/color) - [Overview](/#/color/overview) - [Primary Palette](/#/color/primary) - [Semantic Colors](/#/color/semantic) - [Gradients](/#/color/gradients) - [Typography](/#/typography) - [Typefaces](/#/typography/typefaces) - [Type Scale](/#/typography/scale) - [Guidelines](/#/typography/guidelines) - [Imagery & Motion](/#/imagery) - [Illustration](/#/imagery/illustration) - [Iconography](/#/imagery/iconography) - [Motion](/#/imagery/motion) - [Components](/#/components) - [Buttons](/#/components/buttons) - [Forms](/#/components/forms) - [Cards](/#/components/cards) - [Layout](/#/components/layout) - [Communications](/#/communications) - [Copy & Taglines](/#/communications/copy) - [Social Media](/#/communications/social) - [Email](/#/communications/email) - [Presentations](/#/communications/presentations) - [Resources](/#/resources) - [Downloads](/#/resources/downloads) - [Governance](/#/resources/governance)