For developers building with AI
Your coding agent ships generic UI because it has no design direction. AsciiKit gives it senior-designer judgment right inside Claude Code, Cursor, and Claude.ai: install once, then just describe what you’re building — and it generates the journey, PRD, and DESIGN.md your agent builds from, so it ships the right thing instead of slop.
v0, Lovable, Bolt, Cursor, Claude Code — they generate from the same defaults, so the output converges on the same generic page. The model isn’t the problem. The missing design direction is.
The fix
Not a chat that gives you prose. A governed design engine that produces the three documents your agent actually needs — each one feeding the next.
The screens, the flow between them, and the psychological moments that make or break it — mapped before a line of code.
Problem, users, and requirements with acceptance criteria your agent builds against — the unstated decisions, stated.
A governed design system your agent obeys — type, hierarchy, accessibility, and the anti-slop “don’ts”. The format Cursor and Claude already read.
npm install -g asciikit-mcp && asciikit-mcp init. Paste your free trial key (you get one the moment you sign up) and it wires into Claude Code, Cursor, or Claude.ai.
No commands to learn. Tell your agent “design the onboarding for my app, here’s what it does…” and AsciiKit fires automatically — applying the design engine right inside the conversation.
Journey, PRD, and DESIGN.md — one connected set, grounded in a real design-intelligence engine, not improvised taste. Your agent builds the right thing, on purpose.
Try it — just say this
“Design the onboarding for my budgeting app — it’s for freelancers with irregular income who are anxious about money. Give me the Build Kit.”
No slash command needed. AsciiKit recognizes the design intent and fires automatically — power users can still call /asciikit-kit explicitly.
A generic LLM gives you generic design, because it’s improvising taste per prompt. AsciiKit runs a real design engine: six psychology frameworks, governed constraints that systematically suppress dark patterns and slop, accessibility baked in, and an auditable trace from intent to every design decision. The hardest part of a DESIGN.md — the brand intent, the accessibility rules, the anti-slop “don’ts” — is exactly what the engine produces, instead of what you’d have to hand-author.
Install once, then just talk to your agent. The MCP is free on the trial — subscribe for unlimited use across Claude Code, Cursor, and Claude.ai.