npm package discovery and stats viewer.

Discover Tips

  • General search

    [free text search, go nuts!]

  • Package details

    pkg:[package-name]

  • User packages

    @[username]

Sponsor

Optimize Toolset

I’ve always been into building performant and accessible sites, but lately I’ve been taking it extremely seriously. So much so that I’ve been building a tool to help me optimize and monitor the sites that I build to make sure that I’m making an attempt to offer the best experience to those who visit them. If you’re into performant, accessible and SEO friendly sites, you might like it too! You can check it out at Optimize Toolset.

About

Hi, 👋, I’m Ryan Hefner  and I built this site for me, and you! The goal of this site was to provide an easy way for me to check the stats on my npm packages, both for prioritizing issues and updates, and to give me a little kick in the pants to keep up on stuff.

As I was building it, I realized that I was actually using the tool to build the tool, and figured I might as well put this out there and hopefully others will find it to be a fast and useful way to search and browse npm packages as I have.

If you’re interested in other things I’m working on, follow me on Twitter or check out the open source projects I’ve been publishing on GitHub.

I am also working on a Twitter bot for this site to tweet the most popular, newest, random packages from npm. Please follow that account now and it will start sending out packages soon–ish.

Open Software & Tools

This site wouldn’t be possible without the immense generosity and tireless efforts from the people who make contributions to the world and share their work via open source initiatives. Thank you 🙏

© 2026 – Pkg Stats / Ryan Hefner

@sinups/ai-kit

v0.5.0

Published

Mantine UI kit for agent products: chat, tool calls, diff review, sessions, MCP, agents, skills, permissions, wizards and settings

Readme

AI UI Kit

A UI kit for agent products, built on Mantine. It covers the chat and the screens around it: settings, MCP servers, agents, skills, permissions, hooks, memory, sessions, background tasks and diff review. Every component works in a 360px widget and on a full page, in light and dark color schemes.

Documentation: https://sinups.github.io/ai-kit

Modules

| Module | What it gives you | | --- | --- | | Chat | AgentChat, MessageList, InputBar, Markdown, status and context rows (AgentStatus, ContextUsage, TurnSummary, CompactBoundary), conversation and prompt search | | Tools | Tool call cards (BashTool, EditTool, SearchTool, TodoTool, PlanTool, SubagentTool, McpTool, ...) and ToolApprovalFooter | | Primitives | Wizard, SettingsLayout, MasterDetail, EntityList, CommandPalette, ConfirmDialog, KeyValueEditor, SchemaView, StatusBadge, ShortcutHint | | MCP | Server list and detail, add/edit wizard, import, discovered servers, tool detail | | Agents | Agent list, detail, editor, creation wizard, tool selector | | Skills | Skill catalog, detail, editor, picker | | Permissions | Allow/ask/deny rules panel, rule wizard, permission mode selector | | Hooks | Lifecycle hooks panel and wizard | | Memory | Memory files panel with preview and editor | | Sessions | Session history, preview, export | | Message actions | Edit, retry, rewind, branch, feedback, plan approval | | Tasks | Background tasks panel, task list and detail, agent tree | | Diff | Multi-file diff review with accept and reject | | Model settings | Model, effort, output style, usage, status | | Help | Commands and shortcuts reference | | Elicitation | Forms from MCP elicitation schemas | | Theme | AiKitProvider, AiKitThemeCustomizer, createAiKitTheme, --ae-* tokens | | Launcher | ChatLauncher, mountChatLauncher with Shadow DOM isolation |

Components are controlled: data comes in through props, intent goes out through callbacks. Pure logic (validation, filtering, formatting) is exported next to them.

Installation

npm install @sinups/ai-kit @mantine/core @mantine/hooks @tabler/icons-react

Import the stylesheets once at the root of your app. Either take all kit styles in one file:

import '@mantine/core/styles.css';
import '@sinups/ai-kit/styles.css';

or only the styles of the components you use, like @mantine/core/styles/Button.css. base.css holds the --ae-* tokens and is required once; each component file already includes the styles of the components it renders:

import '@mantine/core/styles.css';
import '@sinups/ai-kit/styles/base.css';
import '@sinups/ai-kit/styles/Wizard.css';
import '@sinups/ai-kit/styles/ChatLauncher.css';

With cascade layers, import @sinups/ai-kit/styles.layer.css (everything inside @layer mantine), or put per-component files into a layer yourself: @import '@sinups/ai-kit/styles/Wizard.css' layer(ai-kit);. JavaScript is tree-shaken the same way: importing Wizard bundles about 4 KB gzip, not the whole kit. Sizes per component are on the Bundle size page.

Usage

import { MantineProvider } from '@mantine/core';
import { AgentChat } from '@sinups/ai-kit';

export function Chat({ messages, status, send, stop }) {
  return (
    <MantineProvider>
      <AgentChat
        messages={messages}
        status={status}
        onSend={({ content }) => send(content)}
        onStop={stop}
        contentWidth={760}
      />
    </MantineProvider>
  );
}

Messages are structurally compatible with UIMessage from the AI SDK, so useChat output can be passed directly. The package does not depend on ai.

Transcript options

  • presentation: 'cards' by default; rowsPresentation shows the flat rows of a terminal client, quietPresentation shows MCP calls as muted lines and folds the work before an answer into one line. Import the value from the package root.
  • approvals: Allow/Deny under any tool call, keyed by toolCallId. A standalone MessageList takes the same map from ToolApprovalsProvider.
  • labels: every text of the chat, one section per component; ChatLabelsProvider does the same for a standalone MessageList.
  • toolCatalog, toolArgs, toolOutputs, locale: tool titles from your MCP servers and formatters for arguments and results.
  • workingRow, toolActivity, animateAppearance, frameBatched: on by default in AgentChat, off in a standalone MessageList. evenSpacing is off in both.
  • Composer: InputBar shows context chips with contextItems, onRemoveContext and onRestoreContext; ModeSelector takes a menu title, a badge per mode and digit shortcuts.

Theming

Without extra setup the kit follows your Mantine primary color, fonts and color scheme. Wrap kit screens in AiKitProvider to apply the kit theme to stock Mantine components inside them and to expose accent, radius, density and color scheme settings:

<MantineProvider theme={appTheme}>
  <AiKitProvider accent="violet" radius="default" density="compact" persistKey="kit-theme">
    <McpSettingsPanel servers={servers} />
    <AiKitThemeCustomizer />
  </AiKitProvider>
</MantineProvider>

See Theming.

Embedding the launcher

In a React app, render ChatLauncher with the chat inside:

<ChatLauncher title="Assistant">
  <AgentChat {...chat} contentWidth="100%" wrapLines />
</ChatLauncher>

On a page you do not control, mount it in a shadow root:

import mantineCss from '@mantine/core/styles.css?inline';
import baseCss from '@sinups/ai-kit/styles/base.css?inline';
import launcherCss from '@sinups/ai-kit/styles/ChatLauncher.css?inline';
import chatCss from '@sinups/ai-kit/styles/AgentChat.css?inline';

const widget = mountChatLauncher(document.getElementById('assistant'), <SupportLauncher />, {
  styles: [mantineCss, baseCss, launcherCss, chatCss],
});

See Embedding the launcher.

Documentation

Repository

| Path | What it is | | --- | --- | | package/ | The npm package @sinups/ai-kit: components, styles, tests and stories | | site/ | The documentation site (Next.js) | | ARCHITECTURE.md | Layers, component contract, file layout | | DESIGN.md | Design tokens, sizes, spacing and theme rules | | CONTRIBUTING.md | Development commands, tests, commit and release rules |

yarn install
yarn storybook    # component playground on http://localhost:8271
yarn dev          # documentation site on http://localhost:4100
yarn test         # dependency check, format check, typecheck, lint and unit tests
yarn build        # build the package into package/dist

Releases are automated with release-please from Conventional Commit titles. See CONTRIBUTING.md.

Credits

This project is a fork of Agent Elements by 21st.dev, released under the MIT License. The original chat component design, behavior, examples, skills and the documentation site come from that project; the implementation was rebuilt on Mantine and the distribution became a single npm package. Build tooling is derived from the Mantine extension template. See NOTICE.

License

MIT