Live data · In-app actions · Human handoff
Answer from live data, help visitors choose, and carry out actions in your app. Start with an FAQ; connect your API when you need more.
Answers in seconds · Voice input · 0 infrastructure · Native Telegram handoff
The host app owns its data and booking flow; FodyChat connects the conversation to them.
The assistant finds pickup and destination, helps select an excursion and collect trip details. The Taxi app controls the booking and its next steps.
Ask for a station by country or genre; the assistant searches the live catalogue and links to station pages. A separate public MCP endpoint exposes station search to compatible clients.
Register allowed actions in your app. The assistant can propose one, ask for confirmation when required, and hand execution back to your code.
How onboarding works — click the button, tell our assistant your email, and your dashboard opens:
In the dashboard you create a project, paste your FAQ or connect sources, and grab the embed snippet.
Start with plain text. Scale up to RAG and custom search tools without changing the widget.
Upload your FAQ, docs or services description. The AI answers from it instantly — no embeddings, no extra cost.
Big content gets chunked and searched semantically. Only the relevant parts reach the prompt.
Define search strategies. The AI calls your backend to answer from live data — prices, availability, catalogues.
Documents, screenshots, photos. The AI understands the content, shows a short summary, and hands the original file to your team.
The AI reads attached documents and answers questions about them — right in the chat.
OCR and description for screenshots and photos, with a one-line summary for the guest.
Guest files are delivered to Telegram, Slack or email — your reply can include files too.
Delete messages, purge files, and choose the model — the controls you expect.
Guests delete their own messages; attached files are removed from storage, not just hidden. Operators moderate too.
Deleting a message removes it from the conversation and from the AI's memory.
Per-tenant model choice: pin a premium model or connect an external OpenAI-compatible / Anthropic endpoint (BYOK).
When the AI can't answer, a real operator steps in — on Telegram, email, Slack and more, replying by text or voice, with auto-translation and the original. All behind the same handoff API.
Colour, position, texts — set them visually in the playground and copy the embed code.
One script tag. The widget streams answers and works on any static site or CMS.
~3 KB upfront · lazy-loads the rest · no dependencies
<script src="https://chat.fodylabs.com/chat-engine.js"></script>
<script>
ChatEngine.mount({
endpoint: 'https://chat.fodylabs.com/api/chat',
title: 'Support chat',
fabIcon: '💬', // text/emoji or an image URL
sendIcon: '/send.svg', // optional: image instead of the "Send" text
theme: { '--chat-primary': '#6366f1' },
payload: { project: 'your-project-id', key: 'pub_your-public-key' }
});
</script>