Stellah.App is an AI customer-support product built in-house: an embeddable chat widget with human agent takeover, a knowledge base and real-time analytics. It had to ship as a working product people could sign up to and run against their own support queues — not a marketing site describing one.
- Client
- Stellah.App
- Industry
- Software (SaaS)
- Scope
- Product design, full-stack build and launch
- Year
- 2025
Results
See it at every size.
- Product
- AI support widget with live human takeover
- Training
- Knowledge base built from uploaded PDFs and documents
- Operations
- Unified inbox, team collaboration, real-time analytics
- Reach
- Multi-language, embeddable on any stack
Scroll inside the frame, or switch to the live site to click through it.
How we got there.
The handover is the feature
Every support bot eventually meets a question it should not answer. Products that hide that moment produce the loop everyone has been trapped in. Live human takeover was built in from the start rather than added as an escape hatch — an agent can step into a conversation mid-thread, with the history already there, so the customer never repeats themselves and never has to ask for a person twice.
Trained on what the business already has
The thing standing between a support team and a working bot is usually the content: nobody has time to write a knowledge base from scratch. Feeding it existing PDFs and documents means a team can go from signup to something that answers real questions in an afternoon, using material they already wrote for humans.
One snippet, then get out of the way
The widget embeds with a single piece of markup on any site, whatever it is built with. That constraint shaped the build: no framework requirement, no plugin per platform, nothing for a customer to integrate beyond a copy and paste. A support tool that needs a developer to install it does not get installed.
Most of our projects start with a call where we say what we would do differently. That call is free and there is no deck.