A chatbot that runs the script and understands the conversation
Most bots do one of two things: they follow your script blindly, or they chat freely and forget to collect leads. With ÖzgürKod, both engines run at once — the flow keeps going while the customer’s side question also gets answered.
Starts at 2.500 ₺ / month
7-day free trial · No setup fee · Upgrade your plan any time
If a message that arrives at 21:00 waits until morning, that customer usually won’t wait until morning.
If your average sale is ₺4,000: the monthly cost of your subscription is not even a single unanswered message.
Two engines in one bot
Flow engine runs the script: greeting, question, branching, lead capture, hand off to a rep. You decide what it asks and where each answer leads — the bot does not improvise or skip steps. That is why the data it collects has the same structure for every customer and can be reported on.
The AI layer, meanwhile, handles understanding: “yeah”, “sure”, “ok” and “👍” all read as yes; if the customer asks for the address mid-flow, it answers from your knowledge base and returns to the step it was on; if someone sends two messages in a row, it replies once, not twice. In other words, it talks like a human without giving up the discipline of the script.
The bot speaks only with the information you give it: what you write in the knowledge base and the prices in your product/service catalog in the CRM. The bans on inventing prices, negotiating and promising work it cannot do are each defined separately in the system instructions. These limits exist to reduce the chance of error; by its nature, AI can still produce a wrong, incomplete or false answer (disclaimer).
Build once, run on six channels
You can target a flow to “all channels” or to a single channel. Whatever the channel, conversations come together in the same inbox.
WhatsApp API
Meta-approved Cloud API number
WhatsApp QR
Your own existing number, via QR
Instagram DM
Direct messages
Messenger
Facebook page messages
Telegram
Bot or your own account
Web Chat
The chat bubble on your site
Drag, connect, validate
You draw the flow; you do not write code. Before you publish, the system checks the flow for you.
12 working step types
Conversation
- Message — provides information, does not wait for a reply
- Question — waits for the answer and saves it to a variable (text, phone, e-mail, multiple choice)
Logic
- Logic — if/else branching
- Business hours check — route here if within hours, there if outside
- Visitor recognition — is this person already registered, or writing for the first time
AI
- AI intent router — tells apart “asking about the price” and “making a complaint”
- Sentiment analysis — routes an angry customer to a separate branch
- Knowledge base — answers an open-ended question from the information you write
Action
- Action — creates a lead, adds a tag, processes the source
- HTTP request — calls an external system (stock, order, your own API)
- Go to flow — hands the conversation over to another flow
- Transfer to a rep — turns the bot off and brings in a human
Ready-made template
Nine industry flows; installed inactive, they go live once you approve.
Auto flow
Describe your business and AI drafts the flow; you edit on top of it.
Validate
Finds broken links, unreachable steps, missing settings and infinite loops before publishing.
Test chat
Try the flow on your own screen without touching the real channel.
Live monitoring
See which customer is at which step right now on the sessions screen.
Clone & translate
Copy a working flow and have it translated into another language with one click.
Import / export
Download the flow as JSON and move it to another account.
Undo
Bring back a flow you deleted by mistake from recently deleted.
Real customers don’t write like that
What makes a bot fail in the field is not the scenario but the person who doesn’t follow it. Each of the items below was written for a real hiccup from the field.
Interrupting question
A customer who asks “do you have shipping?” in the middle of a flow is not handed to a rep. The answer comes from the knowledge base and the bot returns to the same step.
Option matching
To a “Yes / No” question, the customer writes “yep”, “sure”, “of course”, “👍”. They write without Turkish characters. The bot tries to match these to the right option; when it can’t understand, it repeats the question instead of assuming an option.
Rapid-fire messages
If the customer sends “17 pro max” + “what’s the price” as two messages, the bot doesn’t reply twice. It waits briefly, merges the messages and gives a single answer.
Language continuity
When a Turkish-speaking customer throws in “yes please”, the bot doesn’t switch to English and restart the conversation. Language is remembered per person and changes only on a persistent signal.
Zero-token shortcut
Messages like “hello”, “thanks”, “okay”, “👍” never reach the AI. They are answered instantly and use no tokens.
Trigger accuracy
The message “Hi, can I learn the price” starts the price flow. The trigger doesn’t have to be at the start of the message; if several flows match, the most specific one wins.
Name memory
A customer who writes again a week later is not asked their name again; it is remembered from the previous conversation.
Loop protection
The same message reaching a customer over and over because of a miswired flow is blocked at the engine level.
Splitting long responses
Responses that exceed Instagram's character limit are not silently dropped; they are split into meaningful chunks and sent in order.
Channel-specific formatting
Real bold text on WhatsApp, plain text on Instagram. The asterisks the AI generates stay hidden on the customer's screen.
Personalization
You can add variables for name, first name, phone, email, time-of-day greeting, date and time. If a variable's value can't be found, the customer never receives it as raw text.
Response length
You decide how much the bot talks with a single slider: one sentence for a hair salon, a detailed explanation for a real estate agency.
Everything the bot knows, on one screen
You don't put your business into one long text box. You organize the information into categories; on each message the bot reads only the entries relevant to that question, not all of it. It answers more accurately and uses fewer tokens.
- Write down what you don't do, too: “we don't ship”, “we don't serve customers abroad” — so the bot doesn't promise a service you don't offer.
- Scope by channel and language: a rule that applies only on WhatsApp doesn't carry over to Instagram.
- Priority and “always send”: critical information such as business hours is always available in every response.
- Business rules: general behavior rules such as “recommend zero products first” or “never mention a discount for cash payment” apply across all channels at once.
- Bulk import with CSV: export the knowledge base, edit it in Excel, and load it back.
The bot reads prices from your catalog instead of making them up
You don't keep a separate product list for the chatbot. Prices are read from your existing product/service catalog in the CRM — update the price in one place and the bot talks about it accurately.
- Only the product asked about: when a customer asks about a single model, the AI receives the relevant items, not the whole list.
- Installment amounts are calculated: you enter the bank financing rates and the bot calculates the monthly amount. When a price changes, you don't update installments by hand.
- Reading from a poster: upload a photo of your price poster and let the AI read it; nothing is added to the catalog until you approve it.
- No haggling: the bot is instructed not to invent discounts and not to say “we can come down a bit”. The authority to discount stays with you.
Customer: how much is the 17 pro max 256, can I do 12 installments?
Bot: the cash price of the 256 GB model is ₺64,999. Over 12 installments it comes to ₺6,180 a month. Would you like me to check stock availability for color options?
You stay in control
The bot works while you're away, not in your place. You decide when it talks and when it stays quiet.
Business hours conditions
Three modes: always, outside business hours only, during business hours only. A day-by-day calendar, holidays, time zone and the channels the rule applies to are each set separately.
Handoff to a human
When the AI decides the conversation needs a human, an “Needs attention” badge appears in the Message Center and the bot goes quiet for that person.
Manual takeover
The moment a rep writes into the conversation, the bot steps aside. Who stopped it and when is logged; you can turn it back on whenever you like.
Appointment permissions
The bot can create appointments — but never without confirmation. It first asks, “Shall I create it for August 14 at 15:00?”, and only saves it if the customer agrees. You can also keep it fully off.
Automatic leads
Every answer collected in the flow is saved to the customer record: name, phone, budget, preference. The source is tagged as “Chatbot” and shows up separately in reports.
The provider is your choice
Google Gemini, OpenAI, DeepSeek or OpenRouter. It runs with your own API key; your data stays in your isolated database.
Silence watchdog
If the bot goes quiet because of a glitch, you won't hear it from the customer: send errors are logged and unusually long silences are reported.
Access control
You define by role who can access the chatbot screens; the control is enforced not only in the interface but also on the server side.
Your industry's flow is already set up
Pick a template, adjust the steps to fit your business, and publish. Templates are installed inactive — the bot won't talk to customers until you approve it.
Hair transplant / Health tourism
Collects photos, city/country, budget and contact details, then passes hot leads to a sales rep.
Dental clinic
Routes by treatment type, passes pain cases to the emergency line and takes a pre-booking for an appointment.
Real estate
Distinguishes buyers from sellers and collects budget and location to produce leads ready for portfolio matching.
Auto service
Takes vehicle details, complaint type and appointment preference, and passes urgent cases that need a tow truck to a sales rep.
Insurance agency
Branches by policy type, separates renewals from new quotes and hands claims notifications to a person.
Pet hotel / boarding
Asks for animal type, date range and vaccination status, and warns guests upfront if vaccinations are missing.
Dietitian / consulting
Client pre-registration: goal, format (online / in person) and contact details.
General appointment
Industry-independent: transfers to a sales rep during business hours and offers a message option outside them.
Health tourism (English)
For patients abroad who write in English. Installed alongside your Turkish flow.
Add it to your site with one line
One script line is all it takes to add the chat bubble to your site. If you use WordPress, install the ready-made plugin and paste in your key.
- Title, primary color, welcome message and bubble position are set from the panel
- A pre-chat form can ask for name and phone before the chat starts
- Incoming chats land directly in the Message Center; a sales rep takes over from the same screen
- When name and phone are collected, a lead is created automatically and its source is tagged
<script src="https://ozgurkod.com/assets/webchat/widget.js"
data-api-key="YOUR_API_KEY" async></script>Frequently asked questions
How long does setup take?
If you start with one of the industry templates, your first flow is up in a few minutes. The real work is writing the information the bot needs to know: address, business hours, services, and the things you don't do. You can write it in one sitting or bulk upload it via CSV.
What happens if the bot gives a wrong answer?
It can — and we say so upfront. The bot is designed to stay within your knowledge base and hand off what it doesn't know, but because AI is probabilistic, no setup reduces errors to zero. The tools you have: placing a “Transfer to a rep” step anywhere in the flow, taking over the conversation manually and writing the reply yourself, and reviewing what the bot said in session transcripts. For critical matters (binding prices, medical/legal guidance) human verification is essential — see Terms of Use, section 8.
Will customers know they're talking to a bot?
That's up to you: you can introduce the bot in the welcome message or get straight to the point. Tone, forms of address and the closing line are managed in the “Conversation tone” section of the knowledge base; you also choose the reply length.
Can several flows run at the same time?
Yes. You define multiple flows with different triggers; whichever one matches starts. If a customer switches to another topic mid-flow, the conversation can be handed to the relevant flow. You can also set up a welcome flow with an “start on first message” trigger.
Where is my data stored?
Each account's database is isolated; the answers and conversations collected are kept only in your database. AI calls are made with the API key you define; there is no data or model sharing with other accounts.
Is the chatbot a separate product?
No. It comes built into the CRM. The bot writes the leads it collects into the same customer list, the conversation appears in the same inbox, and the appointment lands on the same calendar. You don't switch to a separate panel.
Let’s be honest about AI
AI works probabilistically: despite all the measures we take, the bot may produce information that is wrong, incomplete or not actually true. What the bot says is not a statement by ÖzgürKod; human verification is essential on critical matters.
The safeguards on this page — knowledge base limits, reading prices from the catalog, service scope checks, human handoff — are designed to reduce the chance of error; we do not claim they eliminate it entirely. The user is responsible for what the bot says and for the accuracy of the knowledge base; for matters with real consequences such as binding prices, contracts, payments and medical/legal guidance, human verification is essential.
Full disclaimer — Terms of use, section 8Let the bot handle the message that comes in tonight
Set up the ready-made template, write your business into the knowledge base, choose the channel. Start tomorrow morning with conversations already answered.
View plans