Raven MCP
A design-intelligence server I built: page and layout audits, contrast and tap-target checks, and a taste engine that records my judgment.
What it is
A design-intelligence MCP server I built and use daily: page, layout, and screen audits; contrast, typography, and tap-target checks; and a taste engine that makes my design judgment portable — it records my decisions per surface and audits new work against them.
Why
Design feedback that lives in my head doesn't scale to agents working in parallel. Encoding the audits and the taste as tools means every session judges against the same recorded standard — including sessions I'm not watching.
How I use it
Any design task starts with the right Raven audit instead of eyeballing or hand-rolling a script. The taste tools run the kickoff interview, store the binding, and echo my recorded rules into every audit. Deterministic checks — contrast math, tap-target sizes, occlusion — belong to Raven; my eyes own the final call.
Get it
Raven is published as an MCP server — npx raven-mcp or add it to your client's MCP config. The audits work standing alone; the taste engine gets better the more decisions you record.
Connected
This thread is scoped to Raven MCP. Opening another node starts a new one.
Ask this node