Online support for your site, from an assistant that answers by itself
goya24 reads the pages of your site you pick, and connects to WooCommerce, Zarinpal or your own API — so your agent checks the real order and answers, instead of asking the customer to wait.
The problem
Your team answers the same forty questions every day
«سفارشم کی میرسه؟» is not a hard question. It is a slow one — someone opens the order panel, copies a tracking code, checks the post site, comes back, and types an answer. Forty times a day.
Most tickets are lookups, not problems
Where is my order, is this in stock, did my payment go through. Answering them needs data, not judgement — but a chatbot without that data just apologises.
Chatbots don't know anything real
A bot that can only quote your FAQ will confidently invent a delivery date. That costs more trust than it saves time.
Your customers write at 1am
Peak messaging hours are not office hours. Every night your queue grows, and every morning starts behind.
How it works
Point it at your site. It works out the rest.
Four phases, about ten minutes, and the last one is watching it work.
- 1It scans your site
One URL. We crawl your pages and products, detect your framework and platform, and spot the services you already run — WooCommerce, Sazito, Digify, Shopfa, Zarinpal.
- 2It learns your content
Pages are chunked on headings, embedded, and re-crawled on a schedule so answers stay current. Add docs, PDFs, past tickets or a repository.
- 3It gets tools, not just answers
Connected services become callable tools. It checks the real order, tracks the real parcel, verifies the real transaction — and every write action can require your approval.
- 4You watch it reason
Every reply carries its full trace: what it retrieved, which tools it called, how confident it was. When confidence drops below your line, it fetches a human instead of guessing.
Features
The product, piece by piece
Everything the product does, each on a page of its own, with the real screens.
Pricing
Priced on credits, not seats
Start free. A conversation spends credits for what it actually used, and you top up when you want to. No per-seat bill for a team that grows.
Basic
For trying it on a real site
- 100 AI resolutions per month
- 1 agent, web messenger
- Site crawl and knowledge base
- Community support
Advanced
For a team that answers all day
- 2,000 AI resolutions per month
- Unlimited agents, every channel
- Every integration, plus your own API
- A record of what the agent did on every answer
- Email and Telegram support
Enterprise
For volume, or your own models
- Unlimited resolutions
- Bring your own model keys
- On-premise or private cloud
- SSO, audit log, custom retention
- Named support engineer
Questions
Before you ask
- What stops it making things up?
- It answers only from what retrieval returned, and cites it. Below your confidence threshold it hands off rather than guessing — you set that line, and you can watch every decision in the trace.
- Where does our data go?
- Crawled content and embeddings stay in your workspace. Personal data is masked before anything is logged. On the Scale plan you can run it in your own cloud, or bring your own model keys so text never touches ours.
- Does it work with Iranian services?
- Yes — that is what it was built for. WooCommerce and WordPress for orders and content, Telegram and Bale for chat, email for the inbox — and a Persian agent that reads and writes the way your customers do. More Iranian services join as each is tested with a real account.
- How good is its Persian?
- The whole product is Persian-first, right down to a Persian-tuned embedding model. It replies in whichever language the customer wrote in.
- We already use Intercom or Zendesk.
- Import your resolved tickets as knowledge and the agent starts from how your team already answers. Run it on a slice of traffic first and grow it as you gain confidence.
- How long until it's useful?
- The scan and first index take about ten minutes. Most teams leave it in draft for a day, read the traces, tighten the instructions, then publish.
Your first agent is ten minutes away
Free while you evaluate. Bring your own model keys if you'd rather.