Tutorial Summary
  • Goal: Build a Smart Support Agent that answers docs questions, creates Notion tickets, and schedules Calendly calls
  • Estimated Time: 20–30 minutes
  • Prerequisites: Xpander.ai account, Mintlify documentation, Notion workspace, Calendly account, OpenAI API key
When a developer asks a technical question regarding integration workflows, traditional support systems often fall short. They either provide generic responses or require human agents to handle every query, creating bottlenecks and delays. Frameworks like Agno have made building AI agent prototypes more accessible than ever. However, the true engineering challenge lies in building completely autonomous AI agents that can orchestrate complex workflows and move beyond simple Q&A bots. Xpander.ai addresses this challenge, enabling engineering teams to leap from prototype to production-grade autonomous agents with enterprise-ready infrastructure.

​
🗺️ What We’re Building: Automated Docs Helpdesk

Ever feel stuck answering the same customer questions again and again? No matter how good your documentation is, customers can’t find the right answers to their queries. This leads to repetitive support tickets, increased workload on human agents, and slow response times. Imagine your documentation being part of your support team. With Xpander.ai, we can create an AI agent that lives inside the documentation itself—capable of answering queries accurately and with speed. When unable to find answers, it escalates tickets to Notion for human intervention and provides options to schedule calls via Calendly.

​
✨ Key Features

  • Answer First – Agent’s primary goal is to find answers directly within the documentation
  • Confidence-based responses – Only provides answers when confidence ≥ 0.85
  • Source citation – Always includes documentation URLs for transparency

​
2. Intelligent Escalation

  • Smart ticket creation – Creates Notion tickets when unable to find relevant answers
  • No guesswork – Never provides uncertain responses
  • Human handoff – Seamless transition to human support when needed

​
3. Proactive Scheduling

  • Calendly integration – Offers call scheduling after handling tickets
  • Flexible booking – Users can schedule calls regardless of ticket status
  • Multi-modal support – Combines chat, tickets, and calls for comprehensive support

​
🛠️ Prerequisites

  1. Xpander.ai account – For agent orchestration and deployment
  2. Mintlify documentation – Source of truth for agent responses
  3. Notion workspace – For ticket management and escalation
  4. Calendly account – For call scheduling integration
  5. OpenAI API key – For LLM inference (GPT-4o-mini recommended)
The entire project is available on Xpander.ai platform as a pre-built template. With just a few clicks, you can deploy your version. This template is already in use within the Xpander.ai community.Explore Support Agent Template

​
🧑‍💻 Step-by-Step Implementation

​
Step 1: Setting Up the Agent

Access the Xpander platform at https://app.xpander.ai/ and click “New Agent” at the top right of the workspace.
Agent Creation
Configuration:
  • Agent Name: Mintlify Docs Support Agent
  • Framework: Agno (currently the only framework available on Xpander.ai)
  • LLM Selection: GPT-4o-mini (recommended for speed and efficiency)
  • Knowledge Base: Connect your Mintlify documentation files for RAG (Retrieval-Augmented Generation)
Agent Configuration

​
Step 2: Writing the System Prompt

The system prompt is critical as it determines the core instructions for the agent. Xpander.ai provides an in-built AI to help generate effective prompts. Our Approach: Mission Statement: 
You are the Mintlify agent for xpander.ai. Serve as the primary support channel using only Mintlify docs, OR if confidence is below 0.85, start the fallback process.
Behavior Workflow:
  1. High Confidence (≥ 0.85): Paraphrase the answer concisely and cite the source URL
  2. Low Confidence (< 0.85): Start the fallback process (no guessing allowed)
Smart Support Agent Architecture
Fallback Process:
  • Offer to create a support ticket through Notion
  • After ticket creation or user preference, provide Calendly scheduling option
System Prompt Configuration

​
Step 3: Multi-Agent Coordination

Xpander.ai makes third-party tool integration completely hassle-free through ready-to-use connectors. Configuring Mintlify Search Tool: In the tools tab, configure the Mintlify Search tool with additional instructions for effective documentation querying.
Mintlify Tool Configuration
Attaching Third-Party Agents: Navigate to the Connectors menu and find connectors for Notion and Calendly. Each connector provides ready-made API endpoints for specific actions:
  • Notion: API endpoint for creating support tickets
  • Calendly: API endpoint for scheduling calls
Notion Connector
Calendly Connector
In the Multi-agent tab, attach the third-party agents (Xpander-Notion and Calendly) found in the connectors menu.
Multi-Agent Setup

​
Step 4: Setting Rules and Goals

Encode the most important instructions as Rules and Goals for consistent behavior. Key Rule: 
Always answer user queries solely using xpander.ai's Mintlify-indexed documentation, querying llms.txt and llms-full.txt first.
Key Goal: 
When documentation lacks confident answers, offer to create a Notion support ticket with relevant user query details for further assistance.
Rules and Goals

​
🧪 Agent in Action

​
Scenario 1: Answer Not Found (Escalation Flow)

User Query: 
Hi I need help configuring agent with Agno, is your team available tomorrow morning for a tech session @AI Support engineer
Agent Response Flow:
  1. Initial Assessment: Agent searches documentation but finds no confident answer
  2. Escalation Offer: “I couldn’t locate a relevant entry in our documentation for that specific integration. Would you like to open a support ticket in Notion so our team can assist further?”
  3. Ticket Creation: Creates support ticket #12345 in Notion
  4. Follow-up: “Great, I’ve created a support ticket #12345 for you. Our team will be in touch shortly. Would you also like to schedule a call with our support team to discuss this issue?”
Escalation Flow
Agent Graph
Observe the control-flow in Agent Graph during response generation - this shows how the multi-agent system coordinates between different tools and services.
Results:
  • Notion Ticket Created:
Notion Ticket
  • Calendly Call Scheduled:
Calendly Booking

​
Scenario 2: Answer Found (Direct Response)

When the agent finds a high-confidence answer, the interaction is quick and efficient. It provides a direct response along with the source URL from the documentation.
Direct Answer
Support tickets and their status in Notion:
Notion Dashboard

​
💻 Developer Support Capabilities

While the primary focus is on no-code workflow, Xpander.ai also excels at processing technical queries and providing comprehensive developer support. Example: Technical Integration Query
Developer Query 1
Developer Query 2
Developer Query 3
When developers ask technical questions about integration workflows with Agno, the Mintlify Support Agent provides:
  • Step-by-step descriptions of required integration steps
  • Inline code snippets for specific integration patterns
  • Best practices and troubleshooting guidance
  • Source documentation links for deeper exploration

​
🎯 Final Thoughts

Building the Mintlify Docs Support Agent on Xpander.ai demonstrates how powerful automation can be achieved using no-code platforms. This level of support system automation is a game-changer that significantly reduces human workload while maintaining high-quality assistance. Key Benefits:
  • 24/7 Availability – Solves user queries in seconds, any time of day
  • Smart Escalation – Knows exactly when human intervention is needed
  • Seamless Integration – Connects documentation, ticketing, and scheduling in one workflow
  • No Technical Expertise Required – Built entirely through no-code interfaces
Production Impact:
  • Reduced Support Load – Handles routine documentation queries automatically
  • Faster Response Times – Instant answers for documented topics
  • Better User Experience – Consistent, accurate responses with proper source attribution
  • Efficient Escalation – Human agents only handle complex, undocumented issues
The no-code approach of Xpander.ai enables teams to create sophisticated assistants without backend complexity, infrastructure management, or extensive development resources. Next Steps:
  • Customize the confidence threshold based on your documentation quality
  • Add more connectors for your specific toolstack (Jira, Zendesk, etc.)
  • Implement custom escalation rules for different types of queries
  • Set up analytics to track agent performance and user satisfaction
Ready to transform your support workflow? Start building your intelligent support agent today! 🚀
Quick Start: Use the pre-built template at Xpander.ai Support Agent Template to get started in minutes rather than building from scratch.