GATE's AI support agent is a chat in the corner of your website or app that answers from your own documents, helps signed-in customers get things done, turns complaints into clear bug reports and hands any chat to a person on your team. One script tag adds it, and it loads only on the web addresses you list.
You write the brief and paste one line. It answers from your documents, and you can join any conversation.
Give it a name and a greeting, then write how it should sound, what it may promise and when to bring in a person. Add your help articles, FAQ and policies as text documents, and it treats them as the facts about your product.
One script tag puts a chat button in the corner of every page. It runs in its own layer, apart from your page's styles, and it only loads on the web addresses you list.
On your public site it answers visitors. Inside your product it knows who is signed in, from a token your server signs, and can act for them. Or it takes bug reports.
Register a webhook for each thing it may do, like changing a delivery address. GATE signs every call so your server can check it came from GATE, and every call is logged.
Every conversation shows up in GATE. Anyone on your team can write into a chat or take it over so the agent stops replying, and the customer sees who joined.
It can look at the customer's screen, their recent clicks and the form they're filling in, but only when you switch it on and they say yes.
The customer holds these settings in the chat's own menu. Anything you switch off stays off for everyone.
Every setting starts on Ask each time: the customer sees why the agent wants it and picks Allow once, Always allow, Never or Not now. One privacy switch turns all three off for everyone.
Code enforces them, not the agent.
Each install does one job: answer visitors, help signed-in customers, or take bug reports.
Visitors ask about your plans, delivery or opening hours and get an answer from your own documents, any time of day.
Your server signs a short token with the customer's id, so the agent knows who is signed in and keeps the same conversation on their phone and their laptop.
Each thing it may do is a webhook you register. GATE signs the call, tries again if your server is busy, and logs what came back.
With the customer's OK it reads the form they're stuck on and fills in the fields. They check it and press submit themselves.
In report mode it turns what the customer describes into a report with a title, the details and a severity. Reports land in GATE, and a webhook can pass them on to your helpdesk.
Read any conversation in GATE, write into it, or take it over so the agent stays quiet. Hand it back and the agent picks the chat up again, reading what you wrote.
Tell it how your business talks, what it may promise and when to bring in a person, in plain words. Give it your help articles and policies, and it answers from them.
Meet your agentsAn AI support agent answers a company's customers and takes care of simple requests at any hour. GATE's AI support agent sits in the corner of a website or app, answers from the documents the company gives it, can fill in a form with the customer's OK, turns complaints into bug reports, and a person on the team can take over any conversation.
With one script tag. A company lists the exact web addresses the agent may run on, without wildcards, and pastes the line into its pages, and a chat button appears in the corner. GATE's chat runs in its own layer, apart from the page's styles, and it does not load on any address that is not on the list.
GATE's support agent answers from what the company gives it: a brief on how to talk, what to promise and when to bring in a person, plus help articles, FAQs and policies as text or markdown files of up to 256 KB each. The agent treats those files as the facts about the product, and the brief says what to do when something is not covered.
Yes, for signed-in customers. The company registers a webhook for each action, like changing a delivery address, and GATE's agent calls the right one when the customer asks. GATE signs every call with HMAC-SHA256 and a timestamp, retries up to 3 times when the server is busy, and logs every attempt. On a public site the agent only answers.
Yes. Every conversation shows up in GATE, and anyone on the team can write into it or press Take over, so the agent stops replying and the customer sees that a named person is replying live. Pressing Resume hands the chat back, and GATE's agent reads what the person wrote before it carries on.
The address of the page the customer is on comes with every message. Screenshots, the customer's last 5 clicks and the form they are filling in each need the company to switch them on and the customer to say yes, every time by default. GATE masks password and card fields before anything leaves the customer's browser.
Yes. In report mode GATE's support agent turns what the customer describes into a bug report with a title, the details and a severity of low, medium or high. Reports land in GATE, where the team can read them, and a webhook can send each report on to the company's helpdesk.
GATE's support agent never presses submit on a form and cannot type into password or card fields. It does not alert anyone when a person is needed, so someone has to watch the conversations; it does not learn by itself or write its own memory; and on a public site it has no actions, only answers. Each visitor is capped at 30 messages an hour there.
GATE is invite only while GATE Intelligence onboards a small number of businesses at a time. After the invite, a company names its support agent, writes the greeting and the brief, adds its help articles and policies as files, lists its web addresses and pastes one script tag into its pages. The chat strings come in English and Swedish.
Invite only while we onboard a small number of businesses.