ZChat ZChat 2.0 Self-Hosted Live Chat
Grounded answers

AI that answers from your documents

A general-purpose model knows a great deal about the world and nothing about your refund window, your firmware versions, or the one setting that fixes 40% of your tickets. Give ZChat's AI your own documentation and it answers from that - with no training run, no upload to a vendor, and no waiting for a rebuild when you fix a typo.

Download Free Evaluation AI With Human Handoff

It's just a folder of files

Markdown or plain text, in a folder on your server. No pipeline, no vector database to operate, no import job to babysit. If you can write a help article, you can add to the knowledge base.

Fix a doc, fix the answer

Change a file and the AI uses the new wording - no retraining, no redeploy. When support spots a bad answer, the correction is a text edit, not a project.

Your docs stay yours

Nothing is uploaded to a vendor to be indexed. Run a local model and the documents - and the questions asked about them - never leave your network.

How the retrieval actually works

Your files are split into passages. Each document is broken into chunks so a long manual can contribute just the paragraph that matters instead of drowning the model in everything you've ever written.

The best-matching passages are selected by keyword. When a question arrives, ZChat scores your passages against the words in that question and picks the strongest few. This is keyword matching, not semantic embedding search - we'd rather tell you that than let you assume it understands synonyms it doesn't. In practice it means one thing for you: write your documentation in the words your customers actually use, and the answers get noticeably better.

Those passages go to the model with the question. The AI is told to answer from the supplied material and to say plainly when it can't - which, paired with handoff, turns a gap in your docs into a human conversation rather than a confident invention.

What to put in it

The documents that earn their place are the ones your agents retype most.

Document Why it pays off
Policies Refunds, delivery windows, warranty terms - factual, stable, and asked constantly.
Setup and troubleshooting The step-by-step fixes agents paste every day, in the customer's vocabulary.
Product specifics Model numbers, compatibility, limits - exactly what a general model will otherwise guess at.
Your real FAQ Not the marketing FAQ - the questions your inbox actually receives.
Honest about the edges

Good documents in, good answers out

Grounding makes an AI far more useful, but it doesn't rescue documentation that is out of date or written in language no customer would use. The retrieval is keyword-based, so vocabulary matters. Treat the knowledge base as something your support team maintains, not a box you tick once.

Also feeds

The same passages inform reply suggestions, so your agents' drafts are grounded too.

Where to start

Practical rules in writing docs your AI can use - or edit articles straight from the dashboard.

Do I have to train or fine-tune a model?
No. There is no training step. You put markdown or plain text files in the knowledge base folder, and when a question comes in ZChat finds the relevant passages and hands them to the model along with the question. Editing a document changes the answers immediately.
How does it decide which passages are relevant?
By matching the words in the visitor's question against your documents and taking the best-scoring passages. It is keyword matching, not semantic embedding search, so it rewards documents that use the same vocabulary your customers do. Writing your docs in your customers' words genuinely improves the answers.
What happens when the answer is not in my documents?
The AI is instructed to say so plainly rather than invent something. Combined with handoff, a question your documents do not cover becomes a conversation with a human instead of a confident wrong answer.
Where do the documents live?
In a folder on your own server, as ordinary .md or .txt files. They are not uploaded to a vendor, and with a local model the documents and the questions asked about them never leave your network at all.
Do I have to restart anything after editing a document?
No. The knowledge base is re-read when the files change, so correcting a document is enough to correct the answers. You can also manage the files from the admin area rather than the file system.
Owned customer support software

Deploy live chat on your own terms, not on someone else's pricing model.

ZChat gives you the installable server, web dashboard, website widget, and desktop agent tools in one self-hosted product you buy once and keep. Run it on infrastructure you trust and connect AI only if and how you want it.

Download ZChat See Product Demo

Deployment

Install on Windows or Linux, behind IIS or Nginx, in a VM, or in Docker if that fits your stack.

Commercial model

One-time purchase, perpetual license, and no monthly per-agent bill attached to growth.

AI Flexibility

Use Ollama locally or connect OpenAI and Anthropic with your own provider accounts.