Nutlope/inspo

Real websites for your coding agent to learn design from, searchable over MCP.

What it solves

Agents often lack the design "taste" required to build visually appealing user interfaces. Inspo provides coding agents with a curated library of real-world production websites to study and reference, bridging the gap between functional code and high-quality design.

How it works

It operates as a Model Context Protocol (MCP) server that gives agents access to a massive archive of over 2,300 pages from 832 real sites. The system extracts design tokens—such as color palettes, fonts, spacing scales, and CSS variables—and provides a "fold-by-fold autopsy" of pages. It includes 68 canonical reference components with copy-pasteable JSX. Agents can use a recommend tool to get design suggestions based on a project brief.

Who it’s for

Developers using AI coding agents (such as Claude Code, Cursor, Windsurf, or Zed) who want their generated UIs to look professional and production-ready.

Highlights

  • Extensive Archive: Access to thousands of pages captured on both desktop and mobile.
  • Design Extraction: Automatically extracts semantic palette roles, type ramps, and container widths into a DESIGN.md for every site.
  • Easy Integration: One-command installation for major AI IDEs and agents via npx.
  • Semantic Search: Uses Together AI embeddings to power find_similar and recommendation tools.

Related

  • Project
    zenstory-ai/novel-to-game

    NovelToGame is an open‑source set of AI‑agent skills that turn a full novel into a playable, source‑grounded game (web‑browser or 3D) with full design documentation and a single‑run QA. It works with Claude Code, Codex, or Kimi Code, requires no bundled engine, and records every mechanic’s citation to the original text.

  • Project
    suboss87/FDEOps

    FDEOps is a Node‑JS CLI that supplies a library of 35 “task‑skill” prompts for forward‑deployed engineers. Integrated with an existing AI coding assistant, it structures strategy, architecture and engineering work, stores all customer context in local Markdown files, and runs entirely offline.

  • Project
    OleksandrChekhovskyi/hax

    hax is a lightweight, terminal‑native coding assistant written in C. It talks to many LLM providers (including local llama.cpp and Ollama) via a simple REPL or one‑shot mode, keeping a tiny binary size, low memory use, and full transparency of the request/response transcript.

  • Project
    getsentry/toolkit

    The Sentry Toolkit is an open‑source set of tools (CLI, MCP server, shared libraries) that let AI coding assistants talk to the Sentry error‑tracking platform. It implements a Model‑Context‑Protocol (MCP) service exposing skills such as error/search, inspect, and triage. The server can run remotely (hosted at https://mcp.sentry.dev) or locally via a stdio transport for self‑hosted Sentry. AI‑powered search tools require an LLM provider (OpenAI, Anthropic, Azure OpenAI, or OpenRouter). The repo includes full docs, a local development setup, and testing scripts.