Business-aware AI Agents with
prompt, knowledge, and tools
Build AI Agents inside ChatbotX that answer with the right business context, retrieve knowledge from your documents, run structured AI Functions, and connect to external MCP tools when the conversation needs real action.
System Prompt
Define business context, tone, and response rules.
Knowledge
Let the agent read documents and business references.
AI Functions
Collect fields, trigger flows, and continue naturally.
MCP Servers
Load tools from external systems for real-time operations.
Start in AI Hub, then shape how the agent behaves
The creation flow is straightforward: open AI Hub, create an agent, then define what it should know, how it should speak, and which tools it can use.
Create the agent, then define the operating rules
AI Hub is the control layer for creating, listing, editing, and removing agents. Once an agent exists, the System Prompt becomes the operating brief that tells it how to introduce the business, what tone to use, and which boundaries to follow.
AI Hub workflow
Open AI Hub, choose + Create AI Agent, then manage agents from one list.
Prompt foundation
Set business identity, reply style, and instructions before the agent starts answering.
Ongoing management
Edit, test, or delete agents as the product catalog, policies, and workflows evolve.
Tool-ready design
The prompt works best when it is paired with Knowledge, AI Functions, and MCP tools.
AI Hub
Create, manage, and configure
Agent list
Sales Assistant
Support Agent
Booking Agent
System Prompt
Business intro, tone, and operational rules live here.
Knowledge layer
One reused visual stays here because it still matches the real product story: the agent answering from business documents and controlled reference material.
Turn business documents into usable context
Knowledge gives the agent a place to read business materials before it answers. Upload the files, keep the library organized, then attach the right documents to the right agent so replies stay aligned with your products, policies, and operations.
Library management
Search, review, and remove uploaded files from one list with file name, format, size, and created time.
Agent linking
Attach the right uploaded documents to each agent so it uses the correct business reference set.
Document types that fit naturally
Use business documents and structured files such as PDF, DOCX, TXT, CSV, JSON, slides, spreadsheets, web-style content, and support exports. The point is not the extension list itself. The point is giving the agent grounded material it can actually use.
AI Functions let the agent collect, trigger, and continue
When the conversation needs more than an answer, AI Functions define what data to collect, what output to return, and whether a Flow should run behind the scenes.
Define what the agent should collect and what comes back
Each function tells the AI when it should be used, which parameters it should collect, what raw output will be returned after completion, and whether a Flow ID should run as part of the process.
Function definition
Set a descriptive name, explain the purpose, and give the AI a clear reason to choose it.
Data collection
Specify meaningful parameters like email, city, or date.
Output message
Return a structured result, then let the AI rewrite it into a more natural response.
Trigger Flow ID
Optionally launch a flow to save data, hit a webhook, or route to a human.
Function builder
Structured input for the AI
Purpose
Use this when the user asks for current weather conditions.
Collect
Output
{{data_from_api}}
Optional Flow ID
Save data, call a webhook, or hand the conversation to a human.
Weather lookup
Collect the city, run the function, and answer with current conditions.
Book appointment
Gather date, email, and reason, then return a confirmed outcome.
Connect to human
Trigger a flow with an inbox action so a rep can take over.
Join email list
Collect the email, run the flow, and confirm subscription in the same chat.
MCP Server setup
Bring external tools into the agent
Authentication
Loaded tools
Search product
Check customer purchase history
Create external action request
Connect the agent to advanced external toolsets
MCP Servers extend the agent beyond document lookup and flow triggers. Add the server, choose the authentication method, click Get Tools, then enable that MCP server inside the agent so it can use the loaded toolset in real time.
Server configuration
Create the server with a name, URL, and authentication pattern that matches the external service.
Agent activation
After tools are loaded, select the MCP server from the agent tool settings.
Examples that fit this model
The agent becomes operational when all four layers work together
Prompt sets behavior, Knowledge supplies business context, AI Functions run structured operations, and MCP tools reach external systems when the conversation needs real-time action.
One conversation can move across context, collection, and action
1. Read the request. The System Prompt frames how the agent should behave and which boundaries matter.
2. Check Knowledge. The agent looks through uploaded documents when the answer depends on business material.
3. Collect missing data. If the user request needs a field like city, email, or booking time, the agent can trigger an AI Function and gather it.
4. Run the right tool. The AI Function may launch a flow or return structured data, while MCP can call an external toolset directly.
5. Return a natural reply. The raw output comes back, and the agent rewrites it into a smooth customer-facing answer.
Scenario
Support + action + external data
User asks a specific question
For example: pricing policy, shipping estimate, or order status.
Knowledge checks the business reference
Documents fill the context gap before the agent answers.
AI Function or MCP tool runs
Collect fields, call a flow, or reach an external system.
Agent responds naturally
The user sees one coherent answer, not the toolchain behind it.
Frequently
asked
questions
