An embeddable chat widget, an agent console, a mobile app and an AI knowledge base that reads your existing documents.
- shared packages
- 0shared packages
- client surfaces
- 0client surfaces
- line of embed code
- 0line of embed code
Name withheld under agreement — full walkthrough available on a call
The product ships as a single JavaScript snippet a customer drops on their site — which means the widget has to be tiny, fast, and immune to whatever CSS it lands in. Behind it sits a Node chat server, an agent console, and a Flutter mobile app so agents can answer from a phone.
The AI layer ingests the knowledge a company already has: PDFs and Word documents parsed, chunked and grounded, running on AWS Bedrock. Answers cite what they came from, and hand off to a human when confidence drops rather than inventing something.
Architecturally it is seven shared packages — core, database, SDK, widget, UI, shared and cobrowse — so the widget, console and mobile app cannot drift apart.
Stack
- Node.js
- Next.js
- AWS Bedrock
- AWS SES
- PostgreSQL
- Flutter
Benchmark
Intercom / Tawk.to
Same category, same scope of problem. Built by us, end to end.
Embeddable widget with style isolation
AI knowledge base ingesting PDF and Word documents
Co-browsing for guided support
Agent console with routing and assignment
Flutter mobile app for agents
Email fallback via AWS SES
Want something in this category?
We have built it once already. That means the architecture conversation starts from experience rather than from a whiteboard.
Free 30-minute technical consult. No obligation, no sales script. · We reply within one business day · You own everything we build