Tutorial · Step-by-step

Intermediate

30–45 min

Build an AI Agent with Business Data

Connect customer, subscription, invoice, and usage data to a customer success agent — then build it in Cursor, Claude, ChatGPT, or Mistral.

CSV Export

business-data/

customers.csv
subscriptions.csv
invoices.csv
usage.csv

API Generated

GET/v1/apis/customers

GET/v1/apis/invoices

+ subscriptions, usage, support-tickets

Agent

Customer Success Agent

Answers business questions by calling live REST endpoints—not static files.

Why this matters

AI agents need real data — not hallucinated answers
Business knowledge already exists in spreadsheets
APIs make data accessible to AI systems

Step 1 Step 2 Step 3 Step 4 Step 5 Step 6

Step 1

See the End Result

Show the finished customer success agent. The agent can answer business questions by calling live API Butler endpoints — not by guessing from static files.

Which customers are on the Pro plan?
Which invoices are overdue?
Which accounts have been inactive for 30 days?
Which customers generated the most revenue?

Customer Success Agent

Which customers are on the Pro plan?

Found 12 Pro customers. Acme Corp and Northwind Labs are the highest revenue accounts.

Which invoices are overdue?

Ask about customers, invoices, subscriptions…

Customer Lookup

GET /v1/apis/customers

Subscription Panel

GET /v1/apis/subscriptions

Invoice Panel

GET /v1/apis/invoices

Usage Panel

GET /v1/apis/usage

GET /v1/apis/customersCustomer Lookup
GET /v1/apis/subscriptionsSubscription Insights
GET /v1/apis/invoicesRevenue Overview
GET /v1/apis/usageActivity Analysis
GET /v1/apis/support-ticketsSupport Context

Step 2

Design Your Datasets

Create realistic business datasets. Use shared IDs — customers.csv is the hub. Every other file joins through customer_id.

customers.csv

Central customer table — subscriptions, invoices, support tickets, and usage all join through customer_id.

subscriptions.csv joins via customer_id

invoices.csv joins via customer_id

support_tickets.csv joins via customer_id

usage.csv joins via customer_id

customers.csv

Core customer records — the hub table every other dataset joins to.

idcompany_namecontact_nameemailplancreated_at

csv

      id,company_name,contact_name,email,plan,created_at
1,Acme Corp,Sarah Chen,[email protected],pro,2024-01-10
2,Northwind Labs,James Park,[email protected],enterprise,2024-02-14
3,Blue Ridge Co,Emily Walsh,[email protected],starter,2024-03-22
4,Vertex Systems,Marcus Lee,[email protected],pro,2024-04-05
5,Coastal Analytics,Anna Rivera,[email protected],pro,2024-05-18

subscriptions.csv

Active and cancelled subscriptions linked to customers.

idcustomer_idplanstatusrenewal_date

csv

      id,customer_id,plan,status,renewal_date
1,1,pro,active,2026-07-01
2,2,enterprise,active,2026-12-15
3,3,starter,cancelled,2025-11-30
4,4,pro,active,2026-08-20
5,5,pro,past_due,2026-06-01

invoices.csv

Billing records with payment status and due dates.

idcustomer_idamountstatusdue_date

csv

      id,customer_id,amount,status,due_date
1,1,2499.00,paid,2026-05-01
2,2,8900.00,paid,2026-05-15
3,3,299.00,overdue,2026-04-01
4,4,2499.00,overdue,2026-05-10
5,5,2499.00,paid,2026-05-20
6,1,2499.00,paid,2026-04-01

support_tickets.csv

Customer support context for agent responses.

idcustomer_idprioritystatuscreated_at

csv

      id,customer_id,priority,status,created_at
1,1,high,open,2026-06-01T09:00:00Z
2,2,medium,resolved,2026-05-28T14:30:00Z
3,4,high,open,2026-06-03T11:15:00Z
4,5,low,open,2026-06-05T16:45:00Z
5,3,medium,closed,2026-05-15T10:00:00Z

usage.csv

Product usage metrics for activity and engagement analysis.

idcustomer_idactive_usersapi_requestslast_activity

csv

      id,customer_id,active_users,api_requests,last_activity
1,1,42,185000,2026-06-10
2,2,128,920000,2026-06-11
3,3,3,1200,2026-03-15
4,4,18,45000,2026-06-08
5,5,25,67000,2026-05-02

Step 3

Publish Your APIs

Upload datasets to API Butler and publish endpoints. Each CSV becomes a hosted GET endpoint with pagination and filtering.

Export
CSV Export
Export each business dataset from your spreadsheet tool.
2
Upload Current
Upload to API Butler
One upload per resource — customers, subscriptions, invoices, support tickets, usage.
3
Generate
API Generated
Each CSV becomes a hosted GET endpoint with filtering and pagination.
4
Connect
Endpoints Available
Copy URLs and wire them into your agent or AI coding tool.

CSV Export

business-data/

customers.csv
subscriptions.csv
invoices.csv
usage.csv

API Generated

GET/v1/apis/customers

GET/v1/apis/invoices

+ subscriptions, usage, support-tickets

Agent

Customer Success Agent

Answers business questions by calling live REST endpoints—not static files.

GET/v1/apis/customers

Customer records with plan filtering

GET/v1/apis/subscriptions

Subscription status and renewal dates

GET/v1/apis/invoices

Invoice amounts, status, and due dates

GET/v1/apis/support-tickets

Open tickets by priority and status

GET/v1/apis/usage

Active users, API requests, last activity

bash

      # List customers with pagination
curl "https://api.getapibutler.com/v1/apis/customers?limit=25"

# Filter by plan
curl "https://api.getapibutler.com/v1/apis/customers?filter[plan]=pro"

# Find overdue invoices
curl "https://api.getapibutler.com/v1/apis/invoices?filter[status]=overdue"

# Check inactive accounts (usage last activity)
curl "https://api.getapibutler.com/v1/apis/usage?limit=25"

Sample response — GET /customers?filter[plan]=pro

json

      {
  "data": [
    {
      "id": 1,
      "company_name": "Acme Corp",
      "contact_name": "Sarah Chen",
      "email": "[email protected]",
      "plan": "pro",
      "created_at": "2024-01-10"
    },
    {
      "id": 4,
      "company_name": "Vertex Systems",
      "contact_name": "Marcus Lee",
      "email": "[email protected]",
      "plan": "pro",
      "created_at": "2024-04-05"
    }
  ],
  "meta": {
    "limit": 25,
    "offset": 0,
    "count": 2,
    "total": 12
  }
}

Step 4

Map APIs to Your Agent

Explain which API powers which part of the UI. Readers should immediately understand API → UI relationships — so the agent knows where to fetch data for each question type.

Customer Success Agent

Which customers are on the Pro plan?

Found 12 Pro customers. Acme Corp and Northwind Labs are the highest revenue accounts.

Which invoices are overdue?

Ask about customers, invoices, subscriptions…

Customer Lookup

GET /v1/apis/customers

Subscription Panel

GET /v1/apis/subscriptions

Invoice Panel

GET /v1/apis/invoices

Usage Panel

GET /v1/apis/usage

GET /v1/apis/customersCustomer Lookup
GET /v1/apis/subscriptionsSubscription Insights
GET /v1/apis/invoicesRevenue Overview
GET /v1/apis/usageActivity Analysis
GET /v1/apis/support-ticketsSupport Context

Step 5

Wire Up Frontend & AI

API Butler publishes the data layer — it does not host your agent. You choose where the agent runs, then connect it to the same GET endpoints from Steps 3 and 4.

Where to build your agent

Every platform below uses the same pattern: your agent calls API Butler over HTTPS, receives structured JSON, and answers business questions from live data.

Cursor & Claude Code

Claude (Anthropic)

Where

Claude.ai, Claude API, or Claude Code

How

Register API Butler endpoints as tools. When a user asks about Pro customers or overdue invoices, Claude calls GET /v1/apis/customers or /invoices and answers from structured JSON.

Claude Code integration →

ChatGPT (OpenAI)

Where

Custom GPTs or Assistants API

How

Add API Butler URLs as Actions or function tools. The GPT fetches customer, invoice, and subscription data over HTTPS before responding—no spreadsheet uploads per question.

AI agent prompts →

Mistral

Where

Le Chat, Agents API, or your own app

How

Wire API Butler GET endpoints into Mistral agent workflows. Use the same JSON responses for customer lookup, billing checks, and usage analysis.

Mistral integration →

Tool definition for Claude, ChatGPT, or Mistral agents

json

      // Example tool definition (Claude, ChatGPT, Mistral agents)
{
  "name": "lookup_customers",
  "description": "Find customers by plan, company name, or email",
  "parameters": {
    "type": "object",
    "properties": {
      "plan": {
        "type": "string",
        "description": "Filter by plan: starter, pro, enterprise"
      }
    }
  }
}

// Tool execution → call your API Butler endpoint:
// GET https://api.getapibutler.com/v1/apis/customers?filter[plan]=pro&limit=25
// Response: { "data": [...], "meta": { "total": 12 } }

Build the dashboard in your IDE

For a full customer success agent with chat UI and panels, use Cursor or Claude Code. Paste your endpoint URLs into the prompts below.

User Question

"Which customers are on the Pro plan?"

Agent

Interprets intent and selects the customers API with filter[plan]=pro

API Call

GET /v1/apis/customers?filter[plan]=pro&limit=25

Response

{ "data": [{ "company_name": "Acme Corp", "plan": "pro" }], "meta": { "total": 12 } }

Answer

12 Pro customers found. Acme Corp and Northwind Labs lead by revenue.

Prompt for Cursor

prompt

      Build an AI customer success dashboard that consumes these API Butler REST endpoints.

Endpoints (paste your URLs):
- Customers: [PASTE_CUSTOMERS_ENDPOINT]
- Subscriptions: [PASTE_SUBSCRIPTIONS_ENDPOINT]
- Invoices: [PASTE_INVOICES_ENDPOINT]
- Support Tickets: [PASTE_SUPPORT_TICKETS_ENDPOINT]
- Usage: [PASTE_USAGE_ENDPOINT]

Each endpoint returns { data: [], meta: { limit, offset, count, total } }.

Requirements:
1. AI customer success dashboard with overview metrics — Pro customers, overdue invoices, inactive accounts, top revenue
2. Chat interface — user asks business questions, agent calls the right API and returns structured answers
3. Customer search — lookup by company name or email with pagination
4. Invoice lookup — filter by status (paid, overdue) and sort by amount
5. Subscription lookup — show plan, status, and renewal date per customer
6. Usage analytics — active users, API requests, last activity with inactive account highlighting
7. Reusable API hooks/composables with loading, empty, and error states for each endpoint
8. Responsive layout — chat and panels stack on mobile
9. Reuse existing project conventions; do not add new state libraries unless already present
10. Never hardcode mock rows once endpoints are provided

Prompt for Claude Code

prompt

      Generate a customer success agent dashboard from these API Butler GET endpoints.

Endpoints (paste your URLs):
- Customers: [PASTE_CUSTOMERS_ENDPOINT]
- Subscriptions: [PASTE_SUBSCRIPTIONS_ENDPOINT]
- Invoices: [PASTE_INVOICES_ENDPOINT]
- Support Tickets: [PASTE_SUPPORT_TICKETS_ENDPOINT]
- Usage: [PASTE_USAGE_ENDPOINT]

Phase 1 — Do not write code yet.
Ask me exactly 2–3 focused questions:
- Framework and routing conventions (React, Vue, file structure)
- Which customer fields matter for search, filtering, and agent responses
- How the chat interface should call APIs (direct fetch vs. tool-calling pattern)

Stop and wait for my answers.

Phase 2 — After I answer:
- Generate TypeScript types from API response shapes
- Create a typed API client with reusable fetch utilities for all five endpoints
- Build hooks/composables with loading, error, and empty states
- Implement customer search, invoice lookup, subscription lookup, and usage analytics
- Build the agent chat UI with customer insights panel
- Responsive dashboard shell with navigation
- Keep API keys in env only—never commit secrets
- Do not use placeholder JSON once real endpoints are provided

Step 6

Compare & Ship

API Butler gives you real business data with production-like APIs — not fake placeholder JSON that every developer shares.

JSONPlaceholder

— Fake generic data
— Unrealistic business structures
— No customer–invoice–subscription relationships
— Same placeholder for every developer

API Butler

Real business data from your CSVs
Realistic customer success workflows
Production-like APIs with filtering and pagination
Domain-specific schemas agents can query

Next Steps

APIs for AI Agents

Structured data layer for Cursor, Claude Code, and custom agents.

Turn Business Data into AI-Ready APIs

Transform spreadsheets into endpoints agents can call.

AI Apps with Real Data

Connect AI-generated UIs to live CSV-backed endpoints.

FAQ

AI agent tutorial questions

Where do I build the AI agent?

API Butler provides the data layer — not the agent itself. Build in Cursor or Claude Code for a full dashboard with chat UI (recommended in Step 5). Use Claude API, ChatGPT Custom GPTs/Actions, or Mistral Agents with the same API Butler endpoints registered as tools. All platforms call your GET URLs over HTTPS and answer from structured JSON.

How does an AI agent access API Butler APIs?

Agents call the same REST endpoints your frontend uses. Upload CSV files to API Butler, copy the GET URLs, and pass them to Cursor or Claude Code. The agent fetches structured JSON with predictable fields—customers, invoices, subscriptions—instead of parsing raw spreadsheet files.

Do I need a vector database?

Not for this tutorial. Structured business data—customers, invoices, usage metrics—works best through REST APIs with filtering and pagination. Vector databases help with unstructured documents. For account lookups, revenue questions, and subscription status, API Butler endpoints are the right layer.

Can I use Google Sheets instead of CSV?

Yes. Export your spreadsheet as CSV and upload to API Butler. Google Sheets, Excel, and Airtable exports all work. The tutorial uses CSV files because they are portable, version-controllable, and easy to share with AI coding tools.

Can Cursor generate the frontend automatically?

Yes. Paste your API Butler endpoint URLs into the Cursor or Claude Code prompts in Step 5. For ChatGPT, Claude, or Mistral chat agents, register the same URLs as tools or Actions — the agent fetches live data before answering.

When should I use APIs instead of uploading files directly?

Use APIs when your agent or app needs structured queries—filter by plan, find overdue invoices, check inactive accounts. File uploads work for one-off analysis. APIs give agents repeatable, filterable access to business data without re-parsing spreadsheets on every request.

Ready to build

Upload CSV. Get API. Build your agent.

Start with your five business datasets and have agent-ready endpoints in minutes.

Documentation Cursor integration Claude Code integration Mistral integration All tutorials

Build an AI Agent with Business Data

See the End Result

Design Your Datasets

customers.csv

subscriptions.csv

invoices.csv

support_tickets.csv

usage.csv

Publish Your APIs

CSV Export

Upload to API Butler

API Generated

Endpoints Available

Map APIs to Your Agent

Wire Up Frontend & AI

Where to build your agent

Cursor & Claude Code

Claude (Anthropic)

ChatGPT (OpenAI)

Mistral

Build the dashboard in your IDE

Prompt for Cursor

Prompt for Claude Code

Compare & Ship

APIs for AI Agents

Turn Business Data into AI-Ready APIs

AI Apps with Real Data

AI agent tutorial questions

Where do I build the AI agent?

How does an AI agent access API Butler APIs?

Do I need a vector database?

Can I use Google Sheets instead of CSV?

Can Cursor generate the frontend automatically?

When should I use APIs instead of uploading files directly?

Upload CSV. Get API. Build your agent.