A self-hosted AI chatbot trained on your own docs, matched to your brand, and embedded on your site in one line — for the cost of a Gemini API call.
Built on the stack you already trust
From a 3-person startup to a 50-person support team — point Buildx at your docs and pick the role.
Train it on your help center, FAQ, and refund policy. It answers customers instantly while your team handles the hard ones.
~45% deflection benchmarkGreet visitors, answer pricing questions from your own pricing doc, and hand off qualified leads to your inbox or CRM webhook.
Always on, even weekendsFeed it your catalog or service docs. It explains features, compares plans, and points buyers to the right CTA.
Trained on your catalogDrop in your HR handbook, engineering wiki, or onboarding docs. New hires stop pinging Slack for the same answer twice.
Stays inside your domainPick a theme that matches the use case. Tweak colors and copy in the dashboard whenever you want.
A complete toolkit — visual builder, knowledge training, embed widget, and security baked in.
Every part of the widget is yours to style — colors, fonts, icons, position, welcome copy. No CSS required. Click a swatch to preview.
Flash for speed, Pro for hard questions, Lite to save money. Choose per bot.
Paste a single tag. No npm install, no build step, no developer needed.
<script src="https://your-buildx.app/embed.js" data-bot-id="YOUR_BOT_ID" ></script>
Upload help articles, manuals, FAQs. The bot answers from your content — never makes things up.
Allow only the domains you choose. Stop scrapers and copycats cold.
API keys encrypted at rest with AES-256-GCM. Never reach the browser.
A live health endpoint your team can plug into status pages and alerts.
*Industry benchmark for well-trained AI support bots (Freshworks, 2025)
Commercial AI support tools charge per resolution or per seat. Buildx charges nothing — you pay only for the Gemini API calls you actually make.
Buildx
MIT
Self-hosted & open source. You only pay your Gemini API usage — typically cents per thousand messages with Gemini 2.5 Flash.
Mandatory 50 resolutions/mo minimum. Resolutions billed even on partial answers.
Advanced AI ($50/seat) + Copilot ($50/seat) required for the full feature set.
Sales-focused. Annual contract typical. Custom pricing above the entry tier.
Pricing as of 2026, sourced from each vendor's published rates. Gemini estimate based on Flash @ ~$0.075/1M input tokens for typical short queries.
No prompt engineering. No vector-database setup tutorial. Upload, theme, embed, done.
Buildx handles all the AI plumbing — chunking, embedding, retrieval — so a non-technical teammate can launch a real RAG bot in an afternoon.
Drop in PDFs, Word docs, or plain text from your help center, product manual, pricing page, or wiki. Buildx auto-extracts and learns the content.
Pick colors, fonts, the launcher icon, the welcome message, and the widget position — all from the dashboard. No CSS required.
Paste a single <script> tag into your site,
or share a public chat URL. Lock it to your domain so no one
else can use your bot.
Customers ask questions; the bot replies with answers grounded in your content — 24/7, in dozens of languages, at the cost of a Gemini API call.
Under the hood: each request embeds with Gemini's 768-dim model, queries Qdrant for top-3 cosine matches scoped to the bot, then injects those chunks into the system prompt before the Gemini chat call.
// Embed → retrieve → inject → chat
const queryVec = await getEmbedding(message, key);
const chunks = await qdrant.query(bot._id, queryVec, 3);
const reply = await gemini
.getGenerativeModel({
model: bot.aiModel,
systemInstruction: bot.systemPrompt
+ "\n\nCONTEXT:\n"
+ chunks.map(c => c.content).join("\n---\n"),
})
.startChat({ history })
.sendMessage(message);A non-technical teammate can ship this. Optimized for Vercel — runs entirely on your account.
Clone Buildx to your GitHub. You always own a copy of the code.
Paste the fork URL into our wizard. Set env vars in your browser, push to live.
Paste it in Settings. Stored with bank-grade encryption.
The wizard runs entirely in your browser — we never see your credentials.
Bank-grade encryption, scoped access, rate-limited endpoints. Secure by default — and audit-ready out of the box.
API keys are encrypted at rest with AES-256-GCM — the same standard banks use for transactions. They never leave your server, and never reach the browser.
Signed sessions via NextAuth — no third party in the loop.
Whitelist exactly which sites can load your bot.
Per-IP rate limits keep your Gemini bill predictable.
One toggle disables a bot instantly — no delete needed.
Your Gemini key never touches a customer's browser.
The whole codebase is open — your security team can read every line.
GET /api/health into your status page and uptime
monitors in under a minute.
Buildx never sees your conversations. Every byte stays inside accounts you already control.
Asks a question on your site
Vercel, Render, your VPS
Google Gemini, your key
Qdrant, your docs indexed
MongoDB, settings & bots
Mayon Labs has zero servers in this loop. Every link in the chain is an account you control.
Two minutes from fork to live. MIT licensed. No per-seat bills, no per-message fees.