Lesson 6. Practice: Connect an External Service

Lesson 6. Practice: Connect an External Service#

Goal: connect one external service to your agent and test the integration.

Task: Send a Telegram Notification When a Client Is Added to Google Sheets#

Tools: Google Sheets + Telegram Bot + Zapier (or Make, or n8n).

Step 1. Create a Bot in Telegram#

1. Message @BotFather
2. Send /newbot
3. Enter the bot name (e.g., "My Notifications Bot")
4. Enter the username (e.g., my_notifications_bot)
5. Copy the API Token (e.g., 1234567890:ABCdefGHIjklMNOpqrsTUVwxyz)
6. Send your bot /start (so the bot can message you)

Step 2. Get Your Chat ID#

1. Send your bot any message
2. Open in a browser:

https://api.telegram.org/bot<YOUR_TOKEN>/getUpdates

(replace <YOUR_TOKEN> with the token from step 1)

3. Find "chat":{"id":123456789} in the response — that's your Chat ID
4. Save the Chat ID

Step 3. Create a Spreadsheet in Google Sheets#

Create a "Clients" spreadsheet with columns:

Step 4. Create an Automation in Zapier#

Trigger:

• app: Google Sheets
• event: New Spreadsheet Row
• authorization: grant access to Google Sheets
• select spreadsheet and sheet
• test (Zapier will find the last row)

Action:

• app: Webhooks by Zapier
• action: POST
• URL:

https://api.telegram.org/bot<YOUR_TOKEN>/sendMessage

• Method: POST
• Data (JSON):

{
  "chat_id": "<YOUR_CHAT_ID>",
  "text": "New client: {{Name}} ({{Email}})"
}

(replace <YOUR_CHAT_ID> with your Chat ID from step 2)

• test (you'll receive a message in Telegram)

Turn on the Zap:

• click "Publish"

Step 5. Test#

1. Add a new row to the Google Sheets spreadsheet:

2. Wait 1–2 minutes (Zapier checks the spreadsheet every 1–15 minutes depending on plan)
3. You'll receive a message in Telegram:

New client: Jane Doe (jane@example.com)

Done! The integration works.

Integration Verification Checklist#

• Trigger works (Zapier finds the new row)
• Action runs (message is sent to Telegram)
• Data is passed correctly (name and email from the spreadsheet appear in the message)
• No errors in Zapier log
• Delay is acceptable (1–15 minutes)

Module Practice#

Practice Task 1: Read API Documentation#

Task: read the Telegram Bot API documentation and find how to send an image.

Steps:

1. Open core.telegram.org/bots/api
2. Find the method for sending an image (hint: sendPhoto)
3. Read what parameters are required
4. Find the request example

Answer these questions:

• What method is used to send an image?
• What parameters are required?
• Can you send an image by URL (without uploading the file)?

• Method: sendPhoto
• Required parameters: chat_id, photo
• By URL: Yes, you can pass the image URL in the photo parameter

Practice Task 2: Get an API Key#

Task: get an API key for one of these services (your choice).

Options:

• Telegram Bot API — create a bot via @BotFather
• OpenAI API — sign up at platform.openai.com and create an API key
• Google Sheets API — create a project in Google Cloud Console (harder, but useful)

Checklist:

• API key obtained
• Key stored securely (not in code, not in chat)
• Tested (made a test request)

Practice Task 3: Connect an External Service via Connector#

Task: repeat the example from Lesson 6 (Google Sheets + Telegram Bot + Zapier).

Requirements:

• trigger: new row in Google Sheets
• action: send message to Telegram
• testing: add a row → receive a message

Checklist:

• Bot created, token obtained
• Chat ID obtained
• Spreadsheet created
• Zap created and enabled
• Testing passed (message received)

Practice Task 4: Zapier Practice — Connect an External API#

Task:
Create a chatbot/agent in Zapier that fetches up-to-date information via an external API (e.g., cryptocurrency rate, weather, exchange rates) when the user asks.

What to do:

1. Choose an API to connect (simple, free):

   • CoinGecko API (cryptocurrency rates): https://api.coingecko.com/api/v3/simple/price?ids=bitcoin&vs_currencies=usd
   • OpenWeather API (weather): https://api.openweathermap.org/data/2.5/weather?q=Moscow&appid=YOUR_KEY
   • ExchangeRate API (exchange rates): https://api.exchangerate-api.com/v4/latest/USD

2. Create a chatbot/agent in Zapier:

   • Name: "Info Agent" or "Crypto Rates"

3. Write the prompt:

You are an information assistant.

Your task: when the user asks, fetch up-to-date information via API and respond.

What you can do:
- Show the current Bitcoin rate (use CoinGecko API)
- [Other features if you add them]

Rules:
- When the user asks for the Bitcoin rate, use HTTP Request to the API
- Extract the price from the response
- Reply: "Current Bitcoin rate: $[price] USD"

Format:
- Short answer
- Always include update time: "As of [time]"

4. Connect Webhooks / HTTP Request in Zapier:

   • Skills → Add Skill → HTTP Request (or equivalent)
   • Configure the request:
     • Method: GET
     • URL: https://api.coingecko.com/api/v3/simple/price?ids=bitcoin&vs_currencies=usd
   • Test: Execute → you should get JSON with the price

5. Process the response:

   • If Zapier supports JSON processing in the prompt, extract the value bitcoin.usd
   • If not — use the built-in data processing tool (if available)

6. Test the agent:

   • Write: "What's the Bitcoin rate?"
   • The agent should reply: "Current Bitcoin rate: $45000 USD (as of 10:30)"

Checklist:

• API chosen (CoinGecko or other)
• Chatbot/agent created in Zapier
• Prompt written (with instructions for using HTTP Request)
• HTTP Request configured (URL, GET method)
• Testing passed (agent returns current rate)

What you'll get:
Experience connecting an external API to an agent/chatbot in Zapier. Understanding of how an agent can fetch real-time data from the internet.

Time: 30–40 minutes

Note:
If Zapier doesn't have a direct HTTP Request (or it's hard to set up), use ready-made plugins (e.g., weather or exchange rate plugin if available). The main thing is to understand how external APIs work.

Practice Task 5: Calculate API Usage Cost#

Task: estimate how much using the OpenAI API will cost for your agent.

Given:

• you're planning a support agent for a website
• ~500 conversations per month
• each conversation: user writes ~200 characters (~50 tokens), agent replies ~800 characters (~200 tokens)
• model: GPT-4o mini
• cost (2026): $0.00015 per 1000 input tokens, $0.0006 per 1000 output tokens

Calculation:

1. Input tokens per conversation: 50
2. Output tokens per conversation: 200
3. Cost per conversation:
   • Input: (50 / 1000) × $0.00015 = $0.0000075
   • Output: (200 / 1000) × $0.0006 = $0.00012
   • Total: $0.0001275 (~$0.00013)
4. Cost for 500 conversations: 500 × $0.00013 = $0.065 (~$0.07)

Answer: ~$0.07/month (or ~6 rubles/month at 90 ₽/$).

Your task:

Calculate the cost for:

• 2000 conversations per month
• model GPT-5.2 ($0.01 per 1000 input tokens, $0.03 per 1000 output tokens)
• each conversation: 100 input tokens, 300 output tokens

• Input: (100 / 1000) × $0.01 = $0.001
• Output: (300 / 1000) × $0.03 = $0.009
• Total per conversation: $0.01
• Total for 2000 conversations: 2000 × $0.01 = $20/month

Bonus: what if you use DeepSeek-R1?

• Input: (100 / 1000) × $0.003 = $0.0003
• Output: (300 / 1000) × $0.003 = $0.0009
• Total per conversation: $0.0012
• Total for 2000 conversations: 2000 × $0.0012 = $2.40/month (8x cheaper!)

Artifacts#

After completing the module you'll have:

1. Working Integration#

• Google Sheets + Telegram Bot + Zapier (or equivalent)
• trigger: new row → action: Telegram notification

2. "How to Read API Documentation" Checklist#

• list of steps to find the right endpoint and parameters
• usable for any API

3. "Integration Verification Before Launch" Checklist#

• Trigger works (event is detected)
• Action runs (request is sent)
• Authorization configured (key works)
• Data passed correctly (fields are filled)
• Errors handled (agent doesn't crash on failure)
• Limits considered (not exceeding Rate Limits)
• Cost calculated (we know what we pay)
• Logging enabled (we see what's happening)

4. API Cost Calculator#

Template for calculating API usage cost:

Materials for the Site#

Flashcards#