Brand Lock AI
A persistent, machine-readable cognition layer for autonomous AI systems. Drop a URL and a logo. Receive a runtime brand identity that any AI image, video, copy, or UI generator can follow at runtime — no prompting, no drift.
Reads everywhere
What this isn’t.
Most brand guidelines are PDFs. Static, human-only, passively interpreted, and the moment you hand them to an AI, they collapse into prompts and good intentions.
Brand Lock AI isn’t a PDF. It isn’t a token map, a style guide, a tailwind config, or a prompt template. It’s an operating profile — a runtime cognition layer agents read before every generation.
Twelve engines, one runtime.
Brand Lock AI runs your inputs through twelve extraction engines and emits a single JSON object every model can read. Each engine fills a layer of the cognition stack.
Website Intelligence
Read positioning + lexicon from the live site.
Visual DNA Extraction
Composition, philosophy, and energy from logo + refs.
Typography Detection
Headline, body, and accent type system.
Color Psychology
Hex-precise palette with reasoning per token.
Messaging Analysis
Tone, taglines, CTA behaviour, banned vocabulary.
Brand Archetype Inference
Hero, Outlaw, Sage, Magician — pick the active stance.
Emotional Cognition
8-axis 0–10 energy model for downstream agents.
Market Positioning
Industry, region, and the seat the brand occupies.
Competitive Differentiation
Category clichés the LLM-default would reach for, refused.
Runtime Rule Generator
reject_if conditions agents enforce at generation time.
Asset Recipe Generator
Per-surface recipes for ads, blogs, video, UGC, decks.
Brand Drift Validation
Forbidden patterns + linguistic constraints for self-check.
Energy. Every brand is a 0–10 score on eight axes.
Token systems describe how a brand looks. Energy describes how it feels. Every Brand Lock includes a normalised 8-axis energy model that drives motion pacing, copy intensity, UI density, image aggression, and CTA tone across every downstream surface.
Below: a sample read of an aggressive performance brand. Outputs whose perceived energy diverges by more than ±2 on any axis should be regenerated.
Sample · Athletic performance brand
“Aggressive minimalism with cinematic athletic realism. Discipline as cultural identity.”
Narrative. Every brand has a 3-act arc.
Long-form generation needs structure. Brand Lock AI infers your brand’s starting state, the transformation the brand performs on the audience, and the destination they end up in. Every campaign, video, and landing page snaps to this shape.
§01
Starting state
Hesitant, ordinary effort that nobody notices.
§02
Transformation
Relentless discipline that compounds into excellence.
§03
Destination
Part of an elite culture of winners.
Contexts. Same brand, different intensity per surface.
A brand on LinkedIn is not the same brand on a performance ad. Brand Lock AI emits per-surface tone, energy, pacing, and CTA rules so agents adjust without losing identity.
Tone · Strategic confidence
Energy · Medium · controlled
Tone · Emotion-first
Energy · High · cinematic
Performance ads
Tone · Command-driven
Energy · Extreme · cuts every 0.4s
Investor mode
Tone · Outcome-led
Energy · Restrained
Landing page
Tone · Direct, single-promise
Energy · High at hero, calm below
Blog
Tone · Editorial authority
Energy · Slow, paragraph-led
What you get.
Runtime cognition file
brandlock.json — a validated runtime every model reads at generation time.
Human brand book
.md + a printable HTML brand book with all 18 cognition sections.
Enforceable rules
Forbidden patterns and reject_if rules ship as machine-readable conditions agents check before generation.
Email is asked once you have the brand book in front of you.
§07 · Imprint
Brand Lock AI is a public free tool from the Nagent AI brand-cognition stack.
Nagent AI ships agentic AI systems for enterprise teams. We open-sourced this layer because we believe the next generation of AI tools needs a shared, persistent vocabulary for brand identity — not endless prompt engineering.
v2.0 · 12 engines · 8-axis energy
schema-validated · vision-extracted