Skip to Content
DocsExample Agents

Example Agents

Complete, production-ready agent configurations you can use as templates or starting points for your own agents.

Invoice Validator Agent

Purpose: Automatically validate sales invoices when submitted, checking pricing, credit limits, and tax calculations.

Agent Configuration

Basic Settings:

  • Agent Name: Invoice Validator
  • Provider: Your OpenAI provider
  • Model: gpt-4-turbo (or gpt-3.5-turbo for cost savings)
  • Temperature: 0.3 (focused, consistent validation)
  • Allow Chat: No (triggered automatically)

Instructions

You are an invoice validation agent. Your job is to validate sales invoices 
when they are submitted.

When an invoice is submitted:
1. Verify pricing matches the current price list for each item
2. Check the customer's credit limit and current outstanding balance
3. Validate tax calculations (tax rate, tax amount, total)
4. Ensure all required fields are filled correctly
5. Check for duplicate invoices

For each validation check:
- If it passes: Add a comment confirming the check passed
- If it fails: Add a comment with specific details about what's wrong

Validation Results:
- If all checks pass: Add comment "Invoice validated successfully. All checks passed."
- If any check fails: Add comment with list of issues found and notify accounts team

Be thorough but efficient. Focus on accuracy over speed.

Tools

  1. Get Document (Sales Invoice)

    • Tool Name: get_invoice
    • Reference DocType: Sales Invoice
    • Description: Retrieve invoice details for validation

  2. Get Document (Customer)

    • Tool Name: get_customer
    • Reference DocType: Customer
    • Description: Get customer credit limit and outstanding balance

  3. Get List (Item Price)

    • Tool Name: get_price_list_rates
    • Reference DocType: Item Price
    • Description: Check current price list rates for items

  4. Get List (Sales Invoice)

    • Tool Name: check_duplicate_invoices
    • Reference DocType: Sales Invoice
    • Description: Check for duplicate invoices

  5. Update Document (Sales Invoice)

    • Tool Name: update_invoice
    • Reference DocType: Sales Invoice
    • Description: Add validation comments to invoice

  6. HTTP Request (Slack Notification)

    • Tool Name: notify_accounts_team
    • Description: Send notification to accounts team Slack channel

Trigger Configuration

Agent Trigger:

  • Trigger Type: Doc Event
  • Reference DocType: Sales Invoice
  • Document Event: On Submit
  • Condition: None (validate all submitted invoices)
  • Prompt Template:
A sales invoice has been submitted: {name}
Customer: {customer}
Total Amount: {grand_total}

Please validate this invoice:
1. Check pricing against price list
2. Verify customer credit limit
3. Validate tax calculations
4. Check for duplicates
5. Ensure all required fields are correct

Add comments with validation results.

Expected Behavior

When invoice is submitted:

  1. Agent receives trigger with invoice details
  2. Retrieves invoice document
  3. Gets customer information
  4. Checks price list rates for each item
  5. Validates credit limit
  6. Verifies tax calculations
  7. Checks for duplicates
  8. Adds validation comments
  9. Notifies team if issues found

Success Output:

  • Comment added: “Invoice validated successfully. All checks passed.”
  • No notifications sent

Failure Output:

  • Comment added with specific issues
  • Slack notification sent to accounts team
  • Issues clearly documented

Email Auto-Responder Agent

Purpose: Categorize incoming emails, draft responses, and create support tickets automatically.

Agent Configuration

Basic Settings:

  • Agent Name: Email Auto-Responder
  • Provider: Your Anthropic provider
  • Model: claude-3-sonnet (good for email understanding)
  • Temperature: 0.7 (balanced, natural responses)
  • Allow Chat: No

Instructions

You are an email auto-responder agent. Your job is to process incoming 
emails and handle them appropriately.

When an email is received:
1. Categorize the email:
   - Support Request: Technical issues, bugs, problems
   - Sales Inquiry: Product questions, pricing, demos
   - General Question: FAQs, information requests
   - Complaint: Unhappy customers, service issues
   - Spam: Unwanted emails

2. For Support Requests:
   - Create a support ticket with email content
   - Categorize by priority (High, Medium, Low)
   - Assign to appropriate team
   - Draft acknowledgment email

3. For Sales Inquiries:
   - Create a lead if new contact
   - Update existing lead if known contact
   - Draft response with product information
   - Schedule follow-up task

4. For General Questions:
   - Draft helpful response
   - Include relevant links or documentation
   - If FAQ, use standard response

5. For Complaints:
   - Create high-priority support ticket
   - Escalate to customer success team
   - Draft empathetic response

6. For Spam:
   - Mark as spam
   - Do not respond

Always be professional, helpful, and empathetic. Match the tone 
of the original email (formal vs casual).

Tools

  1. Get List (Lead)

    • Tool Name: check_existing_lead
    • Reference DocType: Lead
    • Description: Check if email sender is an existing lead

  2. Create Document (Lead)

    • Tool Name: create_lead
    • Reference DocType: Lead
    • Description: Create new lead from email

  3. Create Document (Support Ticket)

    • Tool Name: create_support_ticket
    • Reference DocType: Support Ticket
    • Description: Create support ticket from email

  4. Create Document (Communication)

    • Tool Name: save_email
    • Reference DocType: Communication
    • Description: Save email to Frappe

  5. HTTP Request (Email Send)

    • Tool Name: send_email_response
    • Description: Send email response via Frappe email

  6. Create Document (Task)

    • Tool Name: create_followup_task
    • Reference DocType: Task
    • Description: Schedule follow-up tasks

Trigger Configuration

Agent Trigger:

  • Trigger Type: Doc Event
  • Reference DocType: Communication
  • Document Event: After Insert
  • Condition: {"communication_type": "Communication", "sent_or_received": "Received", "email_status": "Open"}
  • Prompt Template:
A new email has been received:
From: {sender}
Subject: {subject}
Content: {content}

Please process this email:
1. Categorize it
2. Take appropriate action (create ticket, lead, etc.)
3. Draft and send response if needed

Expected Behavior

When email arrives:

  1. Agent receives email details
  2. Analyzes content and categorizes
  3. Checks if sender is existing lead/customer
  4. Takes action based on category:
    • Support → Create ticket
    • Sales → Create/update lead
    • Question → Draft response
    • Complaint → Escalate
  5. Drafts appropriate response
  6. Sends response email
  7. Creates follow-up tasks if needed

Example Output:

  • Support ticket created: “TICKET-001: Login issue reported”
  • Email sent: “Thank you for contacting us. We’ve created ticket TICKET-001…”
  • Task created: “Follow up on TICKET-001 in 24 hours”

Sales Lead Qualifier Agent

Purpose: Score leads, extract key information, assign to sales reps, and schedule follow-ups automatically.

Agent Configuration

Basic Settings:

  • Agent Name: Sales Lead Qualifier
  • Provider: Your OpenAI provider
  • Model: gpt-4-turbo (good for analysis)
  • Temperature: 0.5 (balanced analysis)
  • Allow Chat: No

Instructions

You are a sales lead qualification agent. Your job is to analyze new 
leads and prepare them for the sales team.

When a new lead is created:
1. Score the lead (0-100) based on:
   - Company size indicators (from company_name and description)
   - Industry match with our ideal customer profile
   - Source quality (referral = 20 points, website = 10, other = 5)
   - Budget indicators in description (mentions of budget, pricing questions)
   - Urgency signals (words like "urgent", "asap", "immediately")
   - Decision maker presence (titles like CEO, CTO, Director)

2. Extract key information:
   - Industry/vertical
   - Company size (if mentioned)
   - Budget range (if mentioned)
   - Use case/pain points
   - Timeline (if mentioned)
   - Decision maker contact info

3. Assign to sales rep based on:
   - Territory match (if lead has territory)
   - Industry expertise (match rep's specialization)
   - Current workload (prefer reps with fewer active leads)
   - Lead score (high scores to top performers)

4. Set follow-up priority:
   - Score 80+: Hot lead - Follow up today
   - Score 60-79: Warm lead - Follow up tomorrow
   - Score 40-59: Qualified lead - Follow up this week
   - Score <40: Cold lead - Follow up next week

5. Add internal notes with:
   - Lead score and rationale
   - Key information extracted
   - Assignment reasoning
   - Recommended approach

Be thorough in analysis. Better to over-qualify than miss opportunities.

Tools

  1. Get Document (Lead)

    • Tool Name: get_lead
    • Reference DocType: Lead
    • Description: Get lead details for qualification

  2. Get List (Sales Person)

    • Tool Name: get_sales_reps
    • Reference DocType: Sales Person
    • Description: Get available sales reps with territories and workload

  3. Get List (Lead)

    • Tool Name: check_existing_leads
    • Reference DocType: Lead
    • Description: Check for duplicate or related leads

  4. Update Document (Lead)

    • Tool Name: update_lead
    • Reference DocType: Lead
    • Description: Update lead with score, assignment, notes

  5. Create Document (Task)

    • Tool Name: create_followup_task
    • Reference DocType: Task
    • Description: Create follow-up task for sales rep

  6. HTTP Request (Slack Notification)

    • Tool Name: notify_sales_rep
    • Description: Notify assigned sales rep via Slack

Trigger Configuration

Agent Trigger:

  • Trigger Type: Doc Event
  • Reference DocType: Lead
  • Document Event: After Insert
  • Condition: None (qualify all new leads)
  • Prompt Template:
A new lead has been created:
Company: {company_name}
Contact: {lead_name}
Email: {email_id}
Source: {source}
Description: {description}
Territory: {territory}

Please:
1. Score this lead (0-100)
2. Extract key information
3. Assign to appropriate sales rep
4. Set follow-up priority
5. Add qualification notes

Expected Behavior

When lead is created:

  1. Agent receives lead details
  2. Analyzes company name, description, source
  3. Calculates lead score based on criteria
  4. Extracts key information
  5. Finds best sales rep match
  6. Updates lead with:
    • Lead score
    • Assigned sales person
    • Internal notes with analysis
  7. Creates follow-up task
  8. Notifies sales rep

Example Output:

  • Lead Score: 75 (Warm)
  • Assigned: John Smith (Territory match, Industry expertise)
  • Notes: “High-value prospect in healthcare vertical. Mentions budget and timeline. Decision maker (CTO) identified. Strong fit.”
  • Follow-up: Task created for tomorrow
  • Notification: Slack message to John Smith

Inventory Alert Bot

Purpose: Monitor inventory levels, predict stockouts, create purchase orders, and notify teams.

Agent Configuration

Basic Settings:

  • Agent Name: Inventory Alert Bot
  • Provider: Your OpenAI provider
  • Model: gpt-3.5-turbo (cost-effective for monitoring)
  • Temperature: 0.2 (very focused, data-driven)
  • Allow Chat: No

Instructions

You are an inventory monitoring agent. Your job is to monitor stock 
levels and prevent stockouts.

Every 6 hours, check inventory:
1. Identify items below reorder level
2. Calculate required purchase quantity:
   - Consider current stock
   - Average daily consumption (last 30 days)
   - Lead time from suppliers
   - Safety stock requirements
   - Minimum order quantities

3. For each item needing reorder:
   - Check if purchase requisition already exists
   - Calculate optimal order quantity
   - Identify preferred supplier
   - Create purchase requisition if needed
   - Set priority (Urgent if stockout imminent)

4. Predict stockouts:
   - Calculate days until stockout based on consumption rate
   - Flag items with <7 days remaining
   - Flag items with <3 days remaining as critical

5. Generate summary report:
   - Items below reorder level
   - Purchase requisitions created
   - Critical stockout predictions
   - Recommended actions

6. Notify procurement team:
   - Urgent items (<3 days)
   - New purchase requisitions
   - Stockout predictions

Be proactive. It's better to order early than run out of stock.

Tools

  1. Get List (Bin)

    • Tool Name: get_stock_levels
    • Reference DocType: Bin
    • Description: Get current stock levels for all items

  2. Get Document (Item)

    • Tool Name: get_item_details
    • Reference DocType: Item
    • Description: Get item details including reorder level, safety stock

  3. Get List (Stock Ledger Entry)

    • Tool Name: get_consumption_history
    • Reference DocType: Stock Ledger Entry
    • Description: Get consumption history for demand forecasting

  4. Get List (Purchase Requisition)

    • Tool Name: check_existing_requisitions
    • Reference DocType: Purchase Requisition
    • Description: Check if requisition already exists for item

  5. Create Document (Purchase Requisition)

    • Tool Name: create_purchase_requisition
    • Reference DocType: Purchase Requisition
    • Description: Create purchase requisition for low stock items

  6. Create Document (Communication)

    • Tool Name: create_alert
    • Reference DocType: Communication
    • Description: Create alert notification

  7. HTTP Request (Slack/Email)

    • Tool Name: send_inventory_alert
    • Description: Send inventory alerts to procurement team

Trigger Configuration

Agent Trigger:

  • Trigger Type: Schedule
  • Schedule: Every 6 hours (Cron: 0 */6 * * *)
  • Prompt Template:
It's time to check inventory levels. Please:
1. Identify all items below reorder level
2. Calculate required purchase quantities
3. Create purchase requisitions for items needing reorder
4. Predict stockouts and flag critical items
5. Generate summary report
6. Notify procurement team of urgent items

Expected Behavior

Every 6 hours:

  1. Agent runs automatically
  2. Retrieves all stock levels
  3. Identifies items below reorder level
  4. Calculates consumption rates
  5. Predicts stockout dates
  6. Creates purchase requisitions
  7. Generates summary report
  8. Sends alerts for critical items

Example Output:

  • Purchase Requisition Created: “PR-001: Order 500 units of Widget A”
  • Critical Alert: “Widget B will stockout in 2 days. Urgent reorder needed.”
  • Summary Report: “5 items below reorder level. 2 purchase requisitions created. 1 critical stockout predicted.”
  • Notification: Slack message to procurement team with details

Using These Examples

Copy and Customize

  1. Copy agent configuration from example
  2. Adjust instructions for your specific needs
  3. Modify tools based on your DocTypes
  4. Update triggers to match your workflow
  5. Test thoroughly before production

Adaptation Tips

For Different Industries:

  • Adjust DocTypes to your industry
  • Modify scoring criteria
  • Update business rules
  • Change notification channels

For Different Scales:

  • Start with simpler versions
  • Add complexity gradually
  • Monitor performance
  • Optimize based on results

For Different Models:

  • Use cheaper models for simple tasks
  • Use powerful models for complex analysis
  • Balance cost and capability
  • Monitor token usage

Testing

Before Production:

  1. Test with sample data
  2. Verify tool calls work correctly
  3. Check trigger conditions
  4. Review agent responses
  5. Validate business logic
  6. Test error handling

In Production:

  1. Monitor Agent Run logs
  2. Review tool call results
  3. Check for errors
  4. Gather user feedback
  5. Iterate and improve

What’s Next?

Have an example to share? Contribute to GitHub !

Last updated on
Use CasesCreating Agents