Figma plugin · v0.1

Cognitive tokens for conversational AI.

Author the tokens that define how an AI product looks, speaks, and behaves — colour and type, plus timing, prompts, voice, and guardrails.

Install in Figma View source
CogWheel plugin collage showing token categories, editors, and seed set

Design systems already own colour. AI products need more.

Latency, escalation, persona, barge-in, and confidence thresholds usually live as Notion docs or private prompt files. CogWheel treats all of them as tokens: named, typed, owned, and reviewable in one place. Foundation and Semantic sync to Figma Variables. Everything else is authored as structured forms and stored as plugin data.

CogWheel UI with foundation, semantic, AI prompt, and voice token editors

Eleven token categories

Every token has a name, value, owner, description, consumedBy, and status: proposed → review → approved.

Foundation

Colour, type, spacing, size, motion, elevation.

Semantic

Meaning over raw values — state, emphasis, action, chat.

Behavioral

Timing, retry, escalation, turn-taking.

Accessibility

Contrast, motion reduction, cognitive-load limits.

Content

Persona voice, copy templates, terminology.

AI Prompt

Persona instructions, guardrails, model routing.

Conversation

Memory, modality switching, state machine.

Voice / Audio

Voice identity, prosody, barge-in.

Avatar

Expression mapping, gesture, rendering.

Governance

Moderation, handoff, compliance.

Telemetry

Success metrics, confidence, latency budgets.

Get started

Build the plugin, then import the manifest in Figma.

  1. 1
    Clone the repo and build 
    git clone https://github.com/jit170621/Cogwheel.git
cd Cogwheel
npm install
npm run build
  2. 2
    In Figma: Plugins → Development → Import plugin from manifest… and select manifest.json.
  3. 3
    Run Cognitive Tokens. Seed the full set from the home screen, then edit tokens by category.

Full schema: token taxonomy.