Drop in a PDF, a URL, a help-center export — BotCraft reads it, grounds every answer in it, and gives you an embed snippet. No prompts to engineer, no models to host.
An opinionated stack — retrieval, citation, branding, analytics, embed — assembled so you spend the afternoon on copy instead of context windows.
Launch in under five minutes. One line of HTML, anywhere — WordPress, Shopify, Webflow, plain HTML. No CLS, < 28 KB on the wire.
Drop PDFs, Word docs, Notion exports, or a sitemap URL. We chunk, embed, and index — your bot is grounded the moment it's ingested.
GPT-4o, Claude Sonnet, Gemini Pro, or self-hosted Llama. Switch is a setting, not a migration — A/B two models against the same docs.
Name, avatar, color, typography, copy, bubble shape, opening message. On Business, custom CSS — looks like a native part of your site.
Top intents, deflection rate, satisfaction, drop-offs — surfaced as a weekly digest, not a dashboard you forget to open.
Auto-detect and respond in 52 languages, no configuration. Your bot speaks your users' language, fluently.
Watch how the whole thing — upload, train, embed — fits in a single coffee break.
We tried to design BotCraft the way a print editor designs a magazine — by removing what isn't necessary. What's left is the shortest path from your documents to a working chatbot.
Sign up, name it, pick a model, set the persona. Tone defaults to whatever your docs read like.
PDFs, Word docs, Notion exports, sitemaps. The index updates in the background — you don't manage it.
Paste the snippet anywhere. The widget loads async, weighs < 28 KB, won't shift your layout.
Most chatbot platforms are flow-builders with an AI bolted on. We started in the other direction — what's the least configuration that can produce a useful answer? — and the rest fell away.
Answers grounded in your documents, not generic AI guesses. Every reply comes back with a footnote pointing to the paragraph it came from.
AES-256 at rest, TLS 1.3 in transit, SOC 2 Type II. Zero-retention agreements with model providers. Your docs are never used to train public models.
Why spend weeks on integration? A trained chatbot, live on your site, in under five minutes. No coding, no hassle, immediate value.
Top-tier AI models — GPT-4, Claude, Gemini — without the enterprise pricing or hidden fees. Three numbers that mean what they say.
The widget blends with any website's design. With full CSS control on Business, it looks and feels like a natural part of your site.
Turn conversations into strategy. Real-time insights into user questions, chatbot performance, and customer satisfaction.
Start on the free plan forever. Upgrade when traffic asks you to. No contracts, no overage surprises.
For trying things — side projects, prototypes.
For growing businesses with active users.
For large teams, agencies, and real SLAs.
We don't fine-tune a model on your documents. BotCraft uses retrieval augmentation: we index your content as semantic chunks, retrieve the relevant passages at question-time, and pass them to the model as context. The answer comes from your docs — citation included — not from the model's general training.
Yes — name, avatar, accent color, typography, copy, bubble shape, opening message. On Business plans, you can also write custom CSS. Most teams ship a widget that's indistinguishable from native.
OpenAI GPT-4o & GPT-4o-mini, Anthropic Claude Sonnet & Haiku, Google Gemini Pro, and Meta Llama 3.3 for self-hosted setups. Switching is a setting; you can A/B two models against the same docs.
One <script> tag, anywhere before </body>. Works on WordPress, Shopify, Webflow, Squarespace, Framer, and plain HTML. The loader is async, won't block your page, and adds no CLS.
AES-256 at rest, TLS 1.3 in transit, SOC 2 Type II infrastructure. We have zero-retention agreements with the model providers we route through. Your documents are never used to train any model — ours or theirs.
Join thousands of businesses using BotCraft to automate customer support, capture leads, and delight customers — 24/7.