SShortSingh.
Back to feed

shadcn CLI's built-in MCP server lets Claude Code, Cursor and VS Code install UI components

0
·2 views

The shadcn CLI now includes an MCP server that lets AI editors such as Claude Code, Cursor and VS Code browse and install registry components. Developers set it up with a single init command per editor, followed by a restart. The server exposes seven tools for listing, searching and viewing registry items, but it does not write files itself, so the assistant runs the returned shadcn add command. The article also explains how to configure a private registry, such as mebius.design, using a token stored in environment variables. It further covers common troubleshooting steps, including clearing the npx cache when no tools appear.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Log in to join the discussion and vote.

Log in

Related stories

0
ProgrammingDEV Community ·

Vue 3 Pagination Synced to URL Query Parameters, With TypeScript and Behavior-Focused Tests

The article describes how pagination in data-heavy interfaces can be kept in sync with the browser URL by storing the active page in a query parameter, enabling shareable links, bookmarks, and working back/forward navigation. The author migrated the shared RoutePagination component and its unit tests to TypeScript as part of GitLab's Contributors Platform work (MR !3200). The component's two-way binding was modernized using Vue 3.4's defineModel() macro, and unawaited router navigation calls were marked with void to satisfy linting rules. The test suite was also refactored away from inspecting internal component state via wrapper.vm, instead asserting on child props, emitted events, and the current route query. The author argues this black-box approach yields more robust tests that do not break when internal variable names change.

0
ProgrammingDEV Community ·

Developer Shares Lessons on Using a Small Team of Custom Copilot CLI Agents

A developer modernizing their personal website, yonka.dev, moved from an ASP.NET Core Razor Pages app to a static Astro site using GitHub Copilot CLI. Rather than building a large set of specialist agents, they chose to rely mostly on built-in agents and introduced just one custom agent, site-reviewer, to check whether changes met the website's requirements. The reviewer was configured as read-only, reporting evidence-based findings and flagging behavior that needed browser verification. Early testing against the original site surfaced useful gaps and accessibility issues, but the reviewer later over-scoped a smaller task by treating features planned for future milestones as defects. The experience highlights the value of limiting agent specialization to real problems and carefully defining review scope.

0
ProgrammingDEV Community ·

A Practical Stack Checklist for Small-Business Web Apps Using Next.js, Node.js, and SQL or MongoDB

Small-business web apps typically combine forms, dashboards, user roles, and third-party integrations such as payments and email, and the main challenge is choosing a stack a small team can ship quickly and maintain over time. The author recommends first mapping core entities and their relationships: relational data suits SQL, while flexible, document-style data fits MongoDB, and a hybrid approach is common. Next.js allows mixing rendering strategies, using static generation for public pages and per-request fetching for authenticated dashboards. For the API layer, a plain Node.js REST API or Next.js route handlers are favored, with request validation, consistent error shapes, paginated lists, and traceable logs, and roles are modeled as data from the start. Before launch, teams should verify backups with tested restores, error tracking, uptime monitoring, and environment variables kept out of the repository.

0
ProgrammingDEV Community ·

Two Free Checks to Catch LLM Hallucinations Before Paying for a Judge Model

A developer guide argues that hallucination testing should begin with inexpensive checks rather than immediately adopting a paid judge model. It distinguishes factual, contextual, and citation hallucinations, noting that contextual errors are especially costly for support bots and RAG apps. The first proposed test extracts numbers from an answer and flags any that do not appear in the retrieved context, which needs no API key. The second is a self-consistency test that runs the same prompt several times and measures how often the extracted fact agrees, since stable answers tend to indicate knowledge while drifting suggests guessing. The author acknowledges both methods have limits, such as missing number words or misattributed figures, and treats failures as cues for human or judge review rather than proof of error.