Creating an LLM agent with tools

LLM Agents in Cotera can go beyond their pre-trained knowledge by using tools. By giving an agent access to a tool like Google Jobs, you can empower it to fetch real-time, external data and bring it directly into your dataset for analysis. This approach is ideal when you need to enrich your data with live information from the internet.

What You'll Build

In this tutorial, you'll create a job-finding agent that uses the Google Jobs tool to search for recent job postings. The agent will automatically search for jobs, extract relevant details, and return structured data ready for your workflows.

Prerequisites

Before starting, make sure you have:

A Cotera account with access to agents
A dataset where you want to add the agent column (we'll use a list of job titles for this example)

Understanding Agents with Tools

Regular agents can analyze data and make decisions, but they're limited to their training data. Tools extend agents by connecting them to external systems - search engines, APIs, databases, and business applications. This transforms agents from analytical tools into active workers that can fetch real-time data and take actions.

Step 1: Add the Tool to Your Organization

Before creating your agent, you need to add the tool to your organization. This is the key difference from creating regular agents.

Finding and Adding the Tool

Navigate to the Tools section in the left sidebar and click the + button.

You'll see Cotera's extensive library of available tools. For this tutorial, we'll use Google Jobs.

Search for or scroll to Google Jobs
Select the specific function you need: Job Search
Review the tool documentation to understand its inputs and outputs
Click Create a new connection if this is your first time using this tool

Some tools require authentication the first time you use them - you may need to provide an API key, enter credentials, or complete an OAuth flow.

Saving the Tool

After configuring the connection:

Click Save in the upper right corner
Give your tool connection a descriptive name, like "Google Jobs Search"
The tool will now appear in your Tools list in the left sidebar

Your tool is now available to use in any agent you create.

Step 2: Create Your Agent Column

Navigate to the dataset where you want to add the agent.

Click the + button to create a new column
Select LLM from the "New Column" menu
This opens the agent configuration interface

Step 3: Define the Agent's Task

The System Message is where you give your agent its core instructions. This should clearly explain the agent's role and what you want it to accomplish.

For a job-finding agent, you might write:

You are a job-finder agent. Your job is to help me find jobs posted in the last 3 days in the US for the role that I am about to give you.

If you were to test this prompt without tools, the agent would explain that it cannot access real-time job postings because its knowledge is static. This is where tools become essential.

Step 4: Connect the Tool to Your Agent

Now let's give your agent access to the tool you added earlier.

Referencing the Tool

In the System Message field, type @ to see a dropdown of available tools in your organization.

Select Google Jobs Search (or whatever you named your tool connection) from the list.

Instructing the Agent

Update your System Message to tell the agent how to use the tool:

You are a job-finder agent. Your job is to help me find jobs posted in the last 3 days in the US for the role that I am about to give you.

Please use the @Google Jobs Search tool to find current job postings.

The @ reference connects your agent to the tool's capabilities, and the instruction tells the agent when and how to use it.

Step 5: Structure the Output

To ensure your agent returns data in a clean, usable format, define a structured output.

Adding Output Fields

Click the icon to define your output structure.

Add fields for each piece of information you want to extract:

job_title (string): The position title
company (string): The hiring company
location (string): Job location
salary (string): Salary information if available
link_to_apply (string): Application URL

Updating the Prompt

Modify your System Message to instruct the agent to return data in this format:

You are a job-finder agent. Your job is to help me find jobs posted in the last 3 days in the US for the role that I am about to give you.

Please use the @Google Jobs Search tool to find current job postings.

Return your findings in the following format:
- job_title: The position title
- company: The hiring company
- location: Where the job is located
- salary: Salary information if provided
- link_to_apply: URL to the job posting

Step 6: Connect to Your Data

In the Prompt section, link the agent's input to the relevant column in your dataset.

For this example, if your dataset has a column called "Job Titles", you would reference it in the prompt:

Search for: {Job Titles}

This tells the agent to use each value from the Job Titles column as the search query when processing rows.

Step 7: Save Your Agent

Before testing, save your agent configuration:

Click Save in the upper right corner
Give your column a descriptive name, like "Job Finder"
The agent is now saved and ready to test

Step 8: Test the Agent

With your agent saved, use the chat preview to test its behavior.

Running a Test Query

In the chat window, enter a sample job title like "Backend Engineer".

The agent will:

Generate a tool call to Google Jobs Search
Execute the search with your parameters
Return the results in your structured format

You may need to manually approve the tool call by clicking the play button the first time. The agent will then fetch live job data and format it according to your specifications.