An AI chatbot for any website, served by your Next.js app on the official OpenAI SDK.
One route, one script tag, no database.
Live demo · Deploy in 5 minutes · Documentation · npm · Changelog
No Next.js app yet? Deploy the starter to Vercel in one click: your own repository, your OpenAI key, no database and no code to write. askdare.vercel.app/deploy walks you through it and makes the CHAT_SECRET for you. Heroku and the details.
In a Next.js 16 app:
npm install @boldare/askdare zod
npx @boldare/askdare initinit writes the chat route, a bot to start from and .env.local, and adds the chat to your layout. Put your OpenAI key and model in .env.local, run the app, and the chat is there. The bot is a few lines you own:
// askdare.config.ts
import { defineBot } from '@boldare/askdare';
export const bots = [
defineBot({
id: 'support',
model: process.env.ASKDARE_MODEL ?? '',
instructions: 'You answer questions about our product. If you do not know, say so.',
security: { allowedOrigins: ['https://www.example.com'] },
widget: { theme: 'aura', greeting: 'Hi! How can I help?' },
}),
];Getting started takes you from an empty app to a working chat in about 15 minutes.
Or try it first, in five minutes, without your own app.
Knowledge base · Tools · Reply options · Attachments · Voice input · The chat inside your page (React) · Answer ratings · Observability · Chat events · Security
Pick a theme, set your brand's colors, or restyle any part with your own CSS, the way BLDR does. Switch between them on the website, then read Themes and Styling.
The guides are on askdare.vercel.app/docs, with search and a chat that answers from them. The same pages are in docs/, and they ship in the package, so they match the version you installed.
- Start: Getting started · Embedding the chat · Command line
- Features: Knowledge base · Tools · Reply options · Attachments · Voice input · Signed-in visitors
- Look and language: Themes · Styling · Texts · The chat inside your page (React)
- Going live: Deployment · Going-public checklist · Costs · Observability · Chat events · Answer ratings · Privacy
- Reference: Configuration · Troubleshooting · Support matrix · Wire protocol · Custom rate-limit store
- Under the hood: Threat model · Design decisions
- AI coding agents:
llms.txtand theaskdareskill, both in the package (how to use them)
Coming from @boldare/openai-assistant? It stopped working when OpenAI shut down the Assistants API, and askdare is its successor: see Migrating.
| Package | What it is |
|---|---|
@boldare/askdare |
The library: bots, the chat route, the embeddable chat and the CLI |
@boldare/askdare-upstash |
Rate limits and budgets shared by every server instance, in Upstash Redis |
@boldare/askdare-otel |
OpenTelemetry spans for every chat turn, following the GenAI conventions |
Contributing · Security policy · MIT license · Made by Boldare



