Activating and managing Chat AI for Website consists of a few simple steps:

1. Subscribe to Mapaprop Business
2. Activate API Token (this will give you access to the MapapropChatAi API Token)
3. Activate Chat AI from the Website section > Website Console > Mapaprop Chat AI for Website
4. Configure parameters, such as the Chat name and introduction text.
5. Read the Chats from the Messages / Leads Inbox
6. Access the Contact to learn in detail about their Searches, Questions, and Favorite Properties

Website Chat AI

Configure the artificial intelligence assistant that handles inquiries 24/7 on your website.

Overview

Mapaprop Chat AI is an intelligent assistant that integrates into your website to answer visitor inquiries automatically, capture leads, and provide information about properties and services 24 hours a day.

Availability: This feature is exclusive to the Business plan. If you have another plan, you will be offered to upgrade when you try to activate it.

Chat AI features

What does the Chat AI do?

Token consumption

• Each Chat AI message consumes tokens from your monthly AI quota
• Tokens renew automatically every month according to your Business plan
• You can see detailed consumption in AI Services

Activate/Deactivate Chat AI

Activate the chat

1. Click the toggle at the top of the page
2. The system automatically verifies that you have the Business plan
3. If you don't have Business, a modal opens to upgrade your plan
4. If you have Business, the chat is activated immediately on your website

Important: The activation is saved automatically. You don't need to click "Save".

Deactivate the chat

1. Click the toggle to turn it off
2. The widget disappears immediately from your website
3. Previous conversations remain in your Inbox

Note: Deactivating the chat does not affect the tokens already consumed, and previous conversations remain available.

Chat AI configuration

Unified token view

At the top you will see a panel with a link to the unified token view that shows:

• Tokens available in your plan
• Total consumption of all AI services
• Breakdown by service (Chat AI, Instagram AI, others)
• Monthly consumption history

Quick access: Click "View AI Usage" to go to the statistics page.

Assistant name

Defines how the assistant presents itself to visitors.

Field: Free text

Examples:

• "Virtual Assistant"
• "María - AI Assistant"
• "Inquiry Bot"

Recommendation: Use a friendly name that builds trust. You can include the name of your real estate agency.

Welcome message

The first message the assistant sends when a visitor opens the chat.

Field: Text area (up to 4 lines)

Example:

¡Hola! Soy María, asistente virtual de Inmobiliaria Ejemplo.
¿En qué puedo ayudarte hoy? Puedo brindarte información sobre
nuestras propiedades, servicios y responder tus consultas.

Tips:

• Be friendly and professional
• Mention what the assistant can do
• Invite them to ask questions

Maximum tokens per request

Limit of tokens the assistant can use in each individual response.

Range: 200 - 5000 tokens

Default: 2000 tokens

Increment: 100 tokens

What does it mean?

• Low value (200-1000): Shorter and more direct responses, consumes fewer tokens
• Medium value (1000-3000): Balanced responses, ideal for most cases
• High value (3000-5000): More elaborate and detailed responses, consumes more tokens

Recommendation: Start with 2000 tokens and adjust according to the response quality you need.

Maximum tokens per conversation

Total limit of tokens a complete conversation can consume before closing automatically.

Range: 10,000 - 100,000 tokens

Default: 30,000 tokens

Increment: 5,000 tokens

What is it for?

• Prevents a single conversation from consuming all your monthly tokens
• Protects against infinite or malicious conversations
• Controls resource spending per client

Practical example:

• With 30,000 tokens and 2,000 per message → maximum ~15 messages per conversation
• With 50,000 tokens and 2,000 per message → maximum ~25 messages per conversation

Recommendation: 30,000 tokens is enough for useful conversations without consuming excessively.

Chat Types configuration (Dual prompt)

The assistant can behave differently depending on what the visitor is doing. That's why, within "Chat Types configuration" you will find two tabs, each with its own instructions field (prompt):

Mapaprop's general rules are already applied in both cases (do not invent properties, do not give definitive legal advice, always treat as an intermediary). You only add the personality and the information specific to your real estate agency.

Example (Search tab):

Sos el asistente de búsqueda de nuestra inmobiliaria. Tono cercano y
profesional. Resaltás siempre las características clave de cada
propiedad y nuestros 20 años de experiencia en la zona.

The Free chat is configured separately (with its own prompt and a topic system). We explain it in detail further down, in "Conversation modes and how to build your prompts".

💡 Do you want a complete ready-to-copy example? Check out the guide How to configure your Chat AI agent (complete example).

Save changes

Auto-save (toggle)

The activated/deactivated state is saved automatically when you change the toggle. You don't need to do anything else.

Save configuration

To save changes to the name, welcome message, tokens, or prompt:

1. Modify the fields you want to change
2. The "Save" button will be enabled when there are pending changes
3. Click "Save"
4. Confirm the action in the dialog that appears
5. The changes are applied immediately to the website

Note: The changes do not affect conversations in progress, only the new ones that start.

Artificial intelligence model (optional)

Below the configuration you will find the "Artificial intelligence model" card, where you choose which AI model responds in your chat.

Option 1: Mapaprop Default (recommended)

It is the default option. Your chat uses the platform's optimized Mapaprop model, without configuring anything. The consumption is deducted from your monthly token quota as always.

Option 2: Connect your own account (Anthropic or OpenAI)

If you have an API account with Anthropic (Claude) or OpenAI (ChatGPT), you can make your chat respond with the model you choose, using your own API key:

1. In Provider, choose Anthropic or OpenAI
2. In Model, choose one from the list (quality vs speed/cost)
3. In API Key, paste the key from your provider account
4. Click "Test connection" to verify that the key works
5. Save — the system validates the key against the provider before saving it; if it is invalid, nothing is saved

Security of your API key

• The key is saved encrypted and never shown on screen again (you will see "Key configured ✓")
• You can change it or remove it whenever you want
• Removing the key returns your chat to Mapaprop's default model (you have an "Undo" button before saving in case you change your mind)

What happens if my key fails?

If your key stops working (you revoked it, it ran out of credit, etc.), the chat does not break: it automatically responds with Mapaprop's default model until you fix it.

Where to create your API key

Conversation modes and how to build your prompts

This section explains how the assistant converses with your visitors and how to configure it so that it responds exactly what you want.

The visitor's paths

When someone opens the chat on your site, the assistant offers them two options to start:

If you did not activate the free chat, the second option is "📝 Make an inquiry", which leaves a message for your team to answer.

The good thing is that the visitor is not locked into one path. While conversing with the assistant they can:

• Jump to searching properties at any time.
• Leave an inquiry so your team can contact them.

And if they are searching for properties, they can go back to consulting with the assistant. All in the same conversation, without losing the thread.

Free chat: conversations about your topics

The free chat turns the assistant into someone who answers your real estate agency's frequently asked questions (financing, guarantees, requirements, areas, etc.) without a human having to be present.

To activate it:

1. Go to the "Free chat" tab within "Chat Types configuration".
2. Turn on the "Enable free chat" switch.
3. Load the free chat prompt (the assistant's personality and knowledge).
4. Load at least one topic (the allowed subjects).
5. Save.

Attention: the free chat consumes AI tokens for each message, just like the search. That's why it is controlled with topics: the assistant only answers about what you define and refers the rest to your team.

Topic system: define what it can talk about

The topics are the list of subjects the assistant is authorized to respond to in the free chat. They work like a menu: the visitor sees them when starting the conversation and the assistant stays within those limits.

• You load the topics as labels (up to 20).
• Each one is a short subject: "Mortgage loans", "Rental guarantees", "Requirements to rent", etc.
• If a visitor asks something outside the topics, the assistant does not improvise: it offers to leave the inquiry so your team can contact them.

The topics define the scope (what can be talked about). The prompt defines the knowledge (what to answer about each subject). The two work together.

How to build a good free chat prompt

The prompt is where you write the assistant's personality and the real information of your real estate agency. The best way to structure it:

1. One line with who the assistant is and the tone.
2. A block for each topic, with the subject title in uppercase and below it the information you want it to use.

Example (for the topics "Mortgage loans", "Rental guarantees", and "Requirements to rent"):

Sos el asistente de nuestra inmobiliaria. Tono cordial y cercano,
respuestas breves. Si te piden algo puntual de un caso concreto,
orientás en general y derivás al equipo.

CRÉDITOS HIPOTECARIOS:
Trabajamos con líneas de bancos como Nación y Provincia. Financian
hasta el 75% del valor de la propiedad. La aprobación final la define
cada banco. Para una simulación con el sueldo real, derivá al equipo.

GARANTÍAS DE ALQUILER:
Aceptamos garante propietario, recibos de sueldo, seguro de caución
o garantía de monotributista. Cada propiedad puede pedir una
combinación distinta: lo confirma el equipo según el caso.

REQUISITOS PARA ALQUILAR:
Pedimos DNI, comprobante de ingresos, una garantía válida y el
depósito y mes de adelanto del contrato. Para los requisitos exactos
de una propiedad, derivá al equipo.

Tips:

• Use uppercase titles for each topic. The assistant follows them better.
• Do not put data that changes often (exact rates, day's prices). Better "I'll refer you to an advisor for the current number".
• Be specific. The more real information you put, the less the assistant improvises.
• Test your topics. After saving, open the chat on your site and ask about each subject to see how it responds.

Inquiry chat: when a human is preferable

Not everything is solved by AI. When a visitor wants to speak with a person, or asks something outside the topics, the assistant offers to leave an inquiry.

• The inquiry is saved in your Inbox as a contact, with the entire conversation history.
• It does not consume AI tokens (it is a direct message for your team).
• Your team responds from the Inbox when they can.

This way you combine the best of both worlds: the AI handles frequent inquiries 24/7 and your team focuses on the cases that really need human attention.

Frequently asked questions

Why can't I activate the Chat AI?

This feature is exclusive to the Business plan. If you try to activate it with another plan, a modal will open offering you to upgrade your subscription. You can review the available plans in Manage Plan.

How many tokens does each conversation consume?

The consumption varies according to:

• Length of the visitor's questions
• Complexity of the generated responses
• Number of messages in the conversation

On average, a typical conversation consumes between 2,000 and 10,000 tokens. You can monitor consumption in AI Services.

Do the Chat AI conversations appear in my Inbox?

Yes. Each Chat AI conversation is saved automatically in your Inbox as a new contact or added to the history if the contact already exists. The conversations have a special "Chat AI" tag to identify them easily.

What happens if my tokens run out?

When the monthly tokens run out:

• The Chat AI is deactivated automatically on your site
• Visitors will not see the chat widget
• You receive a notification so you know

The tokens renew automatically at the start of each billing period according to your Business plan.

Can I customize the appearance of the widget?

Currently the Chat AI widget uses the colors configured in Site Design. Advanced customization of the widget will be available in future updates.

Does the Chat AI respond in other languages?

The Chat AI is optimized for Argentine Spanish, which is the main language of your market. If a visitor writes in another language, the assistant will try to respond in the same language, but the quality may vary.

How do I improve the quality of the responses?

To improve the Chat AI responses:

1. Complete your profile: Make sure you have updated information in your account
2. Keep your inventory updated: More properties = better responses
3. Configure both prompts: The Search one (how it comments on properties) and the Free chat one (what it answers about your topics). See "Conversation modes and how to build your prompts"
4. Load clear topics: In the free chat, define the allowed subjects and put the information for each one in the prompt
5. Increase tokens per request: Allows more elaborate responses
6. Review conversations: Analyze what inquiries you receive and adjust your prompts and topics

Can I see the Chat AI conversations?

Yes. All conversations are saved in your Inbox and you can:

• See the complete history of each conversation
• Filter by the "Chat AI" tag
• Respond manually if necessary
• See captured contact details

Can the Chat AI schedule visits?

Currently the Chat AI captures the client's intention (request a visit) and saves the data in the Inbox, but does not schedule directly in your calendar. Appointment confirmation must be done manually by contacting the client.

What is the difference between the free chat and leaving an inquiry?

The free chat is handled by the AI: it responds instantly about the topics you configured and consumes tokens. Leaving an inquiry sends a direct message to your team (it stays in the Inbox), does not use AI or tokens, and is answered by a person. The assistant offers to leave an inquiry when the visitor wants to speak with someone or asks something outside your topics.

Does the assistant respond to anything in the free chat?

No. In the free chat the assistant only responds about the topics you loaded, using the information from your free chat prompt. If they ask something outside those subjects, it does not improvise: it offers to leave the inquiry so your team can contact them. This way you maintain control over what the assistant says on behalf of your real estate agency.

Common use cases

Case 1: Real estate agency with high web traffic

Recommended configuration:

• Tokens per request: 2000 (balanced responses)
• Tokens per conversation: 50,000 (long conversations allowed)
• Prompt: Focus on capturing contact details quickly

Benefit: You capture more leads without increasing staff.

Case 2: Specialist in a specific area

Recommended configuration:

• Name: "North Zone Assistant"
• Prompt: "Only north zone properties, mention our knowledge of the neighborhood"
• Tokens per request: 2500 (detailed responses)

Benefit: You reinforce your geographic specialization.

Case 3: Cost control

Recommended configuration:

• Tokens per request: 1500 (concise responses)
• Tokens per conversation: 20,000 (lower limit)
• Frequent monitoring at /config/ai

Benefit: You optimize the use of monthly tokens.

Tips to maximize the Chat AI

• Update properties regularly: The assistant uses your updated inventory
• Configure institutional pages: The assistant uses info from "About Us" and "Services"
• Monitor tokens weekly: Avoid running out of tokens mid-month
• Review conversations: Learn what your clients are looking for and adjust your strategy
• Use the custom prompt: Add your unique value proposition
• Try different configurations: Experiment with tokens to find the ideal balance

Chat Messages

From the Messages / Leads Inbox you can read the messages

When tapping a contact you can see the Consulted Properties, the favorites, and the searches performed

If you access the contact, you will see that each contact has related information