Skip to Content
DocsQuick Start

Quick Start

Get your first agent running in 5 minutes. This guide walks you through creating a working AI agent that can interact with your Frappe data.

Already installed Huf? Great! If not, check the Installation guide first.

What You’ll Build

By the end of this guide, you’ll have a customer support agent that can:

  • Answer questions about customer records
  • Look up customer information on demand
  • Respond naturally in conversation

Step 1: Add an AI Provider

Providers connect Huf to AI services. You’ll need an API key from one:

1. Get an API Key

Pick an AI provider and get your API key:

2. Add Provider in Huf

Navigate to: Desk → Huf → AI Provider

Click New and fill in:

  • Provider Name: OpenAI (or your chosen provider)
  • API Key: Paste your key here
  • Click Save

Your API key is encrypted and stored securely in Frappe.

Step 2: Add an AI Model

Models are the “brains” behind your agent. Let’s add one.

Navigate to: Desk → Huf → AI Model

Click New and fill in:

  • Model Name: gpt-4-turbo (or claude-3-opus, gemini-pro, etc.)
  • Provider: Select the provider you just created
  • Click Save

Use friendly names like gpt-4-turbo. Huf automatically formats them correctly for the provider.

Step 3: Create Your Agent

Now for the fun part—creating your AI agent.

Navigate to: Desk → Huf → Agent

Click New and fill in:

Basic Info

  • Agent Name: Customer Support Assistant
  • Provider: Select your provider
  • Model: Select your model

Instructions

Copy and paste this into the Instructions field:

You are a helpful customer support assistant. Your job is to help users find 
and understand customer information in the system. 

When asked about a customer:
1. Ask for the customer ID or name if not provided
2. Use the get_customer tool to retrieve their information
3. Present the information clearly and helpfully

Always be polite, professional, and concise. If you don't have a tool to do 
something, let the user know what you CAN help with instead.

Settings

  • Temperature: 0.7 (balanced responses)
  • Top P: 1.0
  • Allow Chat: Check this

Save the agent (we’ll add tools next).

Step 4: Give Your Agent a Tool

Tools let agents interact with your Frappe data. Let’s give your agent the ability to look up customers.

Create the Tool Function

Navigate to: Desk → Huf → Agent Tool Function

Click New and fill in:

  • Tool Name: get_customer
  • Description: Get detailed information about a customer by their ID or name
  • Types: Select Get Document
  • Reference DocType: Select Customer
  • Click Save

The description is important—your agent uses it to decide when to use this tool.

Add Tool to Agent

Go back to your agent: Desk → Huf → Agent → Customer Support Assistant

Scroll to the Agent Tool table:

  • Click Add Row
  • Tool Function: Select get_customer
  • Save the agent

Your agent now has the power to look up customer information!

Step 5: Test Your Agent

Time to see it in action!

Navigate to: Desk → Huf → Agent Chat

Select your Customer Support Assistant from the dropdown.

Try these prompts:

Can you show me information about customer CUST-00001?
Tell me about the customer named "ABC Corporation"
What can you help me with?

You should see:

  • The agent asks clarifying questions if needed
  • The agent uses the get_customer tool to fetch data
  • The agent presents information clearly

Behind the scenes: Check Agent Run in the sidebar to see detailed logs of what happened—tokens used, tools called, costs, and more.

You Did It!

You now have a working AI agent that can:

  • Understand natural language
  • Use tools to fetch Frappe data
  • Respond intelligently to user questions

What’s Next?

Learn Core Concepts

  • Interfaces - Understand /huf vs /app and when to use each
  • Agents - Dive deeper into agent configuration
  • Tools - Learn about all available tools
  • Triggers - Make agents run automatically

Expand Your Agent

  • Add more tools: Create, update, delete documents
  • Run on schedule: Automate daily/weekly tasks
  • Trigger on events: Validate documents automatically
  • Custom functions: Call your own Python code

Monitor & Optimize

  • Agent Run - Track usage, costs, and performance
  • Optimize prompts for better results
  • Choose cheaper models for simple tasks

Troubleshooting

Agent not responding?

  • Verify API key has credits
  • Check Agent Run for error messages
  • Ensure model name matches your provider

Tool not working?

  • Check you have permissions for the DocType
  • Verify tool is added to agent’s tool list
  • Make sure Customer records exist in your site

Responses seem off?

  • Refine the instructions to be more specific
  • Lower temperature for more consistent responses
  • Add examples of desired behavior

Need Help? Check the rest of the documentation or visit the Frappe Community  for support.

Last updated on
IntroductionInstallation