hfy is a ready-made backend and chat window for AI conversations. Your app keeps its users and its data. hfy runs the model, calls your tools, remembers every conversation and gives your users a chat that feels like part of your product.
Why are you sending it back?
Why hfy
Getting a model to answer is easy. Turning that into something your customers can rely on is where teams lose their roadmap, and every team rebuilds the same pieces.
hfy answers these once, in a single service you run next to your app, so your team can spend its time on what makes your product different.
Use cases
The same hfy runs all of these. What changes is the instructions you write and the tools you connect.
Customer support
The assistant looks up the order, explains what happened and fixes it on the spot. Refunds wait for a tap, and your team can step into the same conversation whenever they want.
E-commerce
It searches your live catalog, asks for size and budget with a tap instead of a typed reply, and takes a photo of what the customer is looking for.
Team copilot
A separate assistant for your team, with its own instructions and access to your internal tools. Questions about your business get answers from your data, and chores get done.
Bookings & appointments
Clinics, salons, studios, repair shops. The assistant checks your calendar, offers the free slots as choices and confirms the booking, at 9 PM on a Sunday as easily as at noon.
Wherever your users are
The chat lives inside the product your users already open, dressed in your colors. You decide where it goes and what frames it: a floating bubble, a full screen or a side panel.
A help bubble on your store, a panel in your dashboard or a page of its own. A few lines on your page put it there, in your brand's colors.
A screen in your iOS, Android, Flutter or React Native app, under your own header and tabs. Photos come straight from the camera.
Someone starts on their phone during lunch and finishes at their desk. The conversation is the same everywhere, down to the last message, because it lives in hfy rather than in the device.
Electron, Tauri, Windows or macOS apps open it like any other view. Dark mode follows the user's system.
Out of the box
No SDK to learn and no glue to maintain. Choose a model, describe your assistant, and all of this comes with it.
Drop it into a web page or your mobile app. Message bubbles, delivery ticks, formatted replies, photos and documents, dark mode and right-to-left languages. Your users will assume you built it.
Every step is saved the moment it happens. Refreshes, flaky networks, deploys and crashes don't lose a word, and an interrupted reply picks up where it stopped.
OpenAI, OpenRouter, a model on your own hardware, or any other that speaks the common OpenAI format. Switch models without touching your app.
The assistant reaches your systems through MCP, the open standard for AI tools. Every request says who is asking, so your systems decide what that person may see or do.
When details are missing, the assistant sends a short form: choices, yes or no, a line of text. Your user taps through it like any screen in your app.
Refunds, deletions and anything you mark as risky wait for the user's go-ahead. Everything else just runs, without a permission prompt in sight.
A customer, a member of your staff and two assistants can share one conversation. Everyone knows who said what, and each person sees it live.
Long conversations are summarized so they keep fitting the model. Turn on search and the assistant can recall what a customer said weeks ago.
A shopping assistant for customers and an admin assistant for staff, each with its own model, instructions and tools, plus specialists they can hand work to.
Users attach photos and documents straight from the chat. hfy checks them, stores them in S3 or on disk, and passes them to the model.
Health checks, metrics for your monitoring and limits on every conversation. It grows with your traffic, and all it needs beside it is a Postgres database.
The browser only ever holds a short-lived pass to one conversation. API keys, instructions and your server secret never leave your servers.
How it fits
You keep doing what you already do: sign people in. hfy does the rest.
When a signed-in user opens the chat, your backend asks hfy for a short-lived pass to their conversation and picks which assistant answers them.
On a page of your website or a screen in your mobile app. It takes your colors and your users' language, and looks at home in both.
It follows your instructions, uses your tools, asks before risky actions, streams each answer and saves every step, so a refresh or restart changes nothing.
Straight answers
You run it next to your app, on your own infrastructure. Conversations live in your database and nowhere else.
There is no SDK to adopt and nothing to program inside hfy. You describe your assistants in one settings file, and any backend language can talk to it.
Your documents and data stay where they are. Offer them as a search tool and the assistant looks things up when it needs to.
One instance serves one app and all of its users. Building a second app? Run a second copy, and each stays simple.
Get started
hfy ships with llms.txt, a complete guide written for AI coding agents. It covers installing hfy, connecting your model and tools, and placing the chat in your app. Hand it over and review what comes back.
Prefer to do it yourself? The documentation walks through every step.
Read https://alash3al.github.io/hfy/llms.txt and add an hfy chat to our customer dashboard. Customers should be able to ask about their orders.