AI Agent Engineering > Create connections and tools > Tool connections

Tool connections

A tool connection is a connection to an external tool that expands the functionality of your AI agents and allows them to perform more complex tasks.

You can connect to the following tools:

•Agentforce Tool
•Azure AI Search Retriever
•Pinecone Retriever
•Python Code
•REST API Tool
•S3 File Parser and Loader
•Salesforce Tool
•Serper Search
•Spider Web Tool

You can use the Python Code and REST API tools to connect to your organization's Informatica Intelligent Cloud Services assets like Application Integration processes, Data Integration tasks, and Data Quality rules.

Agentforce Tool

The Agentforce Tool automates and manages tasks using intelligent agents. It helps your organization streamline its workflows and reduce manual work.

The Agentforce Tool is commonly used in IT operations, customer service automation, or data management scenarios.

Example: Automating IT support ticket routing

Your IT support team receives hundreds of help desk tickets daily. Manually sorting, prioritizing, and routing these tickets to the correct teams causes delays and errors.

Using the Agentforce Tool, you create intelligent agents (or bots) that automatically categorize incoming tickets based on content like software, hardware, or network issues.

The agents route tickets to the right support teams without human intervention, and they can escalate urgent issues based on predefined rules. You only need to monitor agent activity and adjust workflows as needed.

Agentforce Tool connection details

You need to configure the Agentforce Tool before you can use it in an AI agent.

The following table describes the properties of an Agentforce Tool connection:

Property	Description
Name	Name of the connection to the Agentforce Tool.
Location	Project or folder to save your assets. By default, assets are saved to the Default project.
Tool Type	Type of tool being configured.
Description	Optional. Description of the connection to the tool.
Salesforce API Host	Base URL of the Salesforce instance used for API requests by Agentforce Tool. It ensures API requests get redirected to the correct Salesforce organization (production, sandbox, or custom domain). The following examples are typical values: - Production or developer environments: https://api.salesforce.com - Sandbox or testing environments: https://test.salesforce.com
Salesforce Org Domain	Custom domain URL of your Salesforce organization used by Agentforce Tool to establish a connection. It identifies the exact Salesforce instance, including any custom or "My Domain" URLs associated with your organization. A typical value might be the following URL: https://yourcompany.my.salesforce.com
Client ID	ID assigned to your application or integration when it was registered on the Agentforce platform.
Client Secret	Confidential credential associated with the Client ID to authenticate your application securely.
Agent ID	Specific identifier of the agent or workflow you want to invoke or manage.

Azure AI Search Retriever Tool

The Azure AI Search Retriever Tool connects to Azure Cognitive Search to perform search queries enhanced with AI capabilities. It can process large datasets using natural language processing and semantic search to return relevant results.

The Azure AI Search Retriever Tool enhances traditional search using natural language processing and semantic search to return the most relevant information from large datasets or document collections.

Example: Enterprise knowledge base search

Your organization maintains a large collection of documents, manuals, FAQs, and policies that employees frequently access. Using the Azure AI Search Retriever Tool, the employees can search using natural language queries to find the most relevant documents quickly. They can apply filters or ranking to prioritize certain documents types or only retrieve recently updated material.

Azure AI Search Retriever connection details

You need to configure the Azure AI Search Retriever connection before you can use it in an AI agent.

The following table describes the properties of an Azure AI Search Retriever Tool connection:

Property	Description
Name	Name of the connection to the Azure AI Search Retriever Tool.
Location	Project or folder to save your assets. By default, assets are saved to the Default project.
Tool Type	Type of tool being configured.
Description	Optional. Description of the connection to the tool.
Service Name	Name of the Azure Cognitive Search service instance you want to connect to. The Service Name is a string corresponding to your Azure search service (for example, "mysearchservice"), which forms part of the service URL like the following example: https://mysearchservice.search.windows.net
API Key	Admin or query API key for authentication with the Azure Cognitive Search service.
Index Name	Name of the search index you want to query.
Model	Embedding model to use for the AI search.
Text Content Field Name	Field within your Azure Cognitive Search index that contains the main textual content you want to retrieve or search over.
Content Vector Field Name	Field within an Azure AI Search index that stores vector embeddings representing your content. These vector embeddings are numerical representations of text or other data, enabling semantic search capabilities. This allows Azure AI Search to go beyond keyword matching and retrieve results based on semantic meaning.
Semantic Configuration	Optional. Name of the configuration to use for semantic search, should a semantic search option be specified in the Search Type field.
Search Type	Type of search to use: - Similarity. Retrieves documents based on their semantic similarity to a user's query rather than exact keyword matches. This search type can handle synonyms, paraphrases, and semantic nuances to deliver more meaningful results. - Similarity Score Threshold. Similar to the Similarity search type, except only documents that score above the Score Threshold value are returned. This method returns only high-confidence matches, improving precision. - Hybrid. Combines both keyword-based and semantic search methods to deliver more comprehensive and relevant results. This method can capture exact keyword matches that might otherwise be missed by semantic searches. - Hybrid Score Threshold. Same as Hybrid, except only documents that score above the Score Threshold value are returned. - Semantic Hybrid. A more advanced version of Hybrid. Uses keywords but also ranks documents semantically to produce the search results. This method is best for searches involving complex queries or conversational language. - Semantic Hybrid Score Threshold. Same as Semantic Hybrid, except only documents that score above the Score Threshold value are returned.
Top K Results	Number of highest-ranked search results to be returned for a query. Enter a positive integer value up to 1000. A smaller K focuses on the most relevant documents, while a larger K can provide more breadth but may include less relevant items.
Score Threshold	Minimum relevance score a document must have to be included in the search results. Enter a decimal value between 0.0 and 1.0. This value is used when you choose any of the search types that involve a threshold. Note: When the tool performs a Hybrid Score Threshold or Semantic Hybrid Score Threshold search, the generated score values are typically lower than 0.04. A higher threshold means only very relevant results are returned, while a lower threshold includes more, possibly less relevant results.

MCP Tool

Use the MCP Tool to allow your AI agent to connect to an MCP server and perform actions on a data source or API that's exposed as an MCP endpoint. The MCP Tool uses the MCP endpoint to discover the actions that can be performed.

The MCP Tool can connect to MCP servers that use MCP protocol version 2025-06-18 and streamable HTTP transport.

To configure the MCP Tool, you perform the following steps:

1Configure the connection details including the endpoint URL, authentication method, and access token, if required.
2Fetch the list of callable resources that are exposed by the MCP server.

The resources that are exposed by the server are also called tools. They define the actions that can be performed using the MCP server. For example, an MCP server might return resources named send_email and send_email_with_attachment which enable your AI agent to send an email or send an email with an attachment.

3Enable the specific actions that your AI agent is allowed to take.

For example, if you want your AI agent to be able to send an email but not an email with attachments, enable the send_email action but not the send_email_with_attachment action.

To retrieve the list of resources from the server, click Fetch Actions and select a Secure Agent. The MCP Tool uses the Secure Agent to communicate with the server and retrieve the list. There's no limit to the number of resources that the MCP Tool can fetch from the server.

When you retrieve the list of resources, all the actions that can be performed are disabled by default. Enable the actions that you want your AI agent to use.

You can fetch and re-fetch the list of resources at any time. If a resource is deprecated and removed from the MCP server, it's removed from the list of actions the next time you click Fetch Actions. Save the MCP Tool to permanently remove the action.

Most actions have parameters. For example, the send_email action might have parameters for the email recipient, the subject line, and the body. To see the parameters associated with an action, click > to expand the action.

To use the MCP Tool, add it to an agent block in the agent flow. When the MCP Tool is called by an agent block, the agent determines the parameter values. If you expand the tool within the agent block, the Value field for each parameter shows "Agent Determines Value." For more information about adding tools to an agent block, see Inline subagents and tools.

You can't add an MCP Tool to a tool block in an agent flow.

Example: Using the MCP Tool to identify a software code change

You are creating an AI agent to assist with code development. Your company stores its source code in a GitHub repository. During testing, your QA team notices that a field is missing from the user interface. You want your AI agent to identify the exact commit that caused this change.

You create an MCP Tool that connects to a third-party Git MCP server. You retrieve the actions from the server and enable the following actions:

•git_clone to create a copy of the target GitHub repository.
•git_log to query the commit history based on criteria like the author, date, or message.
•git_diff to examine the changes introduced by a particular commit and assess their potential impact.

You add the MCP Tool to an agent block within your agent flow. When you run the AI agent, it uses the MCP Tool to identify the specific commit. The agent can then report this to the development team.

MCP Tool connection details

You need to configure the MCP Tool before you can use it in an agent block.

The following table describes the properties of a MCP Tool connection:

Property	Description
Name	Name of the connection to the MCP Tool.
Location	Project or folder to save your assets. By default, assets are saved to the Default project.
Tool Type	Type of tool being configured.
Description	Description of the actions this tool will perform in the AI agent.
URL	Endpoint URL of the MCP server that your AI agent will connect to, for example: https://api.example.com/mcp/
Timeout	Maximum amount of time, in seconds, that the AI agent waits for a response from the MCP server.
Authentication	Authentication method used to connect to the MCP server to secure communication and control access. Choose one of the following methods: - No Authentication. Select this option when the MCP server uses no authentication, for example, for a public or open access server or when the MCP server operates in a tightly controlled and isolated network. - Token Authentication. Select this option when the MCP server uses an access token for authentication.
Token Prefix	HTTP authentication scheme used by the MCP server to authorize requests. This field is displayed when the token authentication method is selected. You can select a value, enter a custom prefix, or omit the prefix: - Choose Authorization:Bearer or Authorization:Token based on the server requirements. - Enter a custom prefix when the server requires or permits a custom prefix. Use the following format: <header key>:<token prefix> - Omit the prefix when the server automatically handles the prefix or accepts a raw token.
Token	Access token. This field is displayed when the token authentication method is selected.

Pinecone Retriever tool

The Pinecone Retriever tool enhances search and retrieval tasks by efficiently finding semantically similar items within large datasets using vector similarity techniques.

Note: You can't use the Pinecone Retriever tool if your organization uses an outgoing proxy server to connect to the internet.

In a Pinecone database, all documents, text, images, and other data are converted into vector embeddings. When a user performs a query, the Pinecone Retriever tool converts the query into a vector embedding using AI models. The tool then queries Pinecone's vector index to return the top relevant documents or data snippets based on the vector embeddings.

The Pinecone Retriever tool can handle large volumes of data and can return more meaningful results compared to traditional keyword matching. This tool scales to accommodate growing datasets.

Example: Semantic document search

Your organization uses the Pinecone Retriever to power an internal search engine that allows employees to find relevant documents based on the meaning of their queries rather than exact keyword matches. The documents are vectorized and stored in Pinecone, and queries are converted to vectors to retrieve the closest matches quickly.

Pinecone Retriever tool connection details

You need to configure a Pinecone Retriever tool before you can use it in an AI agent.

The following table describes the properties of a Pinecone Retriever tool connection:

Property	Description
Name	Name of the connection to the Pinecone Retriever tool.
Location	Project or folder to save your assets. By default, assets are saved to the Default project.
Tool Type	Type of tool being configured.
Description	Optional. Description of the connection to the tool.
Index Name	Name of the specific vector index within Pinecone containing your document embeddings for search.
API Key	Authentication token that grants access to your Pinecone account and services.
Model	AI or embedding model used to convert input queries and documents into vector embeddings that Pinecone can index and search against. Different models can vary in retrieval accuracy, relevance, and performance.
Namespace	Optional. Logical grouping within a Pinecone index to segment data. You should specify a namespace if hosting multiple datasets in one index. If a namespace is not specified, the default or global namespace is used.
Text Key	Optional. Name of the document field that contains the textual content to be retrieved and displayed when a vector search returns matching documents.
Search Type	Type of search to use: - Similarity. Retrieves documents based on how closely a user's query matches the embedded vectors in the Pinecone database. This search method can handle semantic meanings to deliver more relevant results, especially when the same keywords aren't shared between queries and documents. - Similarity Score Threshold. Similar to the similarity search type, except only documents that score above the Score Threshold value are returned. This method returns only high-confidence matches, improving precision. - Max Marginal Relevance. Retrieves a set of documents that are both relevant yet diverse. This algorithm reduces redundant or overly similar results and is ideal for exploratory searches where your goal is to get results that are both relevant and varied.
Top K Results	Optional. Specifies the number of top results to retrieve from the Pinecone vector index for each query. Enter a positive integer value up to 1000. A higher value includes more candidates but may include less relevant results. A lower value improves precision but could potentially miss some relevant documents. If a Top K Results value is not specified, the tool uses a default value defined by the tool or the Pinecone service.
Score Threshold	Optional. Threshold value to return results. Enter a decimal value between 0.0 and 1.0. Documents above this threshold are returned, while those that score below the threshold are ignored. If no threshold is specified, then all results up to the Top K Results value are returned.
Fetch K	Optional. Number of top candidate vectors to retrieve initially from Pinecone before applying any further filtering. Enter a positive integer value up to 1000. The Fetch K value is typically greater than the Top K Results value, because it fetches all candidates from Pinecone before any filtering. The Top K Results value is the number of final results returned to the user after filtering. If a Fetch K value is not specified, then the tool uses a default set in the Pinecone configuration. This may impact the breadth or precision of the search results.

Python Code tool

The Python Code tool lets you run Python code snippets within your AI agent.

Use the Python Code tool to create your own custom tool to use in your AI agent. The following examples list the kinds of tools you can create:

•PowerPoint presentations
•Complex analytics algorithms
•Connections to Informatica Intelligent Cloud Services assets such as Application Integration processes, Data Integration tasks, and Data Quality rules.
•Connections to a target that isn't currently configured for the AI agent.

To configure the Python Code tool, configure the connection details and then enter your Python code on the Code tab. Define required parameters in the Input Parameters and Output Parameters tabs.

The Python interpreter and some Python libraries are included in the Secure Agent installation. If you receive "missing library file" errors when the AI agent runs your Python code, you'll need to install additional Python libraries on the Secure Agent machine. For more information, see External library installation for custom Python code.

Security notice

The Python Code tool grants your AI Agent powerful capabilities to execute code dynamically. While all execution occurs within a secure, sandboxed environment on the Secure Agent, it's critical that you, as the agent's designer and administrator, understand the associated risks and your responsibilities related to execution of pre-configured code. Pre-configured code is the code that you explicitly provide to the Python Code tool during the configuration of your agent.

As the person who configures the AI Agent, you are responsible for implementing the necessary safeguards. Audit all provided scripts, and thoroughly vet all the Python code that you use. Ensure that the code is free from potentially unsafe operations, including, but not limited to, the following operations:

•Outbound network connections to untrusted endpoints
•Unauthorized file system access or data connections
•Execution of shell commands or other system-level operations

By configuring and enabling the Python Code tool for an AI Agent, you acknowledge and accept responsibility for the code it executes. Always treat the code run by an AI Agent with the same security diligence you apply to any other code within your enterprise environment.

Example: Data transformation in ETL pipelines

A data engineer in your organization uses the Python Code tool within an ETL pipeline to clean raw sales data by removing duplicates, correcting date formats, and calculating key metrics before loading the data into a data warehouse.

Python Code connection details

You need to configure the Python Code tool before you can use it in an AI agent.

The following table describes the properties of a Python Code tool connection:

Property	Description
Name	Name of the connection to the Python Code tool.
Location	Project or folder to save your assets. By default, assets are saved to the Default project.
Tool Type	Type of tool being configured. The Python Code tool is a custom tool.
Description	Description of the actions this Python Code tool will perform in the AI agent. Since the Python Code tool can perform tasks autonomously, a description is required to communicate the purpose and ensure correct usage. Also, regulatory or organizational policies may require a description.

You also need to enter the Python code block and specify input and output parameters.

Python code

The Python Code tool requires a code block to run.

In the Code tab, enter the actual Python code or script to be executed, either as a code block or file reference. For more information, see Python Code connection details.

Specify input or output parameters for your code block using the Input Parameters or Output Parameters tabs. For more information, see Python Code input and output parameters.

Python Code input and output parameters

Define input and output parameters for your Python code block.

To create input and output parameters:

1Select either Input Parameters or Output Parameters from your Python Code connection page. For more information, see Python Code connection details.
2Click the Add Row icon and enter the parameter details:

Property	Description
Name	Name of the parameter.
Type	Parameter type: - String - Integer - Float - Boolean
Description	Description of the parameter.
Default Value	Default value for the parameter.

3Click anywhere outside the row to save your entry.

REST API Tool

The REST API Tool lets your AI agent communicate with external systems or services using REST over HTTP. Use this tool to send requests to and receive responses from RESTful web services.

Using the REST API Tool, you can interact with any exposed REST API. For example, use this tool to refer to a REST API exposed by an Application Integration process or Data Integration task.

To use the REST API Tool, enter the connection details and then define your operations and parameters. Optionally, you can enter request headers, the request body, and the authentication method.

Example: Integrating with a payment gateway

Your organization uses the REST API tool to send payment requests to a third-party payment processor’s API. The tool handles constructing the HTTP requests, authenticating, and processing the responses to confirm payment status.

REST API Tool connection details

You need to configure the REST API Tool before you can use it in an AI agent.

The following table describes the properties of a REST API Tool connection:

Property	Description
Name	Name of the connection to the REST API Tool.
Location	Project or folder to save your assets. By default, assets are saved to the Default project.
Tool Type	Type of tool being configured.
Description	Description of the actions this REST API Tool will perform in the AI agent. Since the REST API Tool can perform tasks autonomously, a description is required to communicate the purpose and ensure correct usage. Also, regulatory or organizational policies may require a description.

You also need to enter the Python code and specify input and output parameters.

REST operation

The Operation tab defines the operation to be performed by the REST API Tool. You can define one operation per REST API Tool.

To define an operation for the REST API tool, perform the following steps:

1Select the Operation tab from within the REST API Tool connection.
2Click the Add Row icon and enter the operation details:

Property	Description
Operation	Name of the operation, for example "Get token".
Verb	HTTP verb for the REST operation: - GET - POST - PUT - DELETE - PATCH The verb is also known as the "HTTP method."
Host	Base URL or domain name of the server where the API is hosted. The host forms the root portion of the full URL that clients use to make requests to the API.
Endpoint	URL route where the operation is accessible, for example: /users/
Description	Description of the operation.

REST API Tool parameters

REST API connection parameters configure how a client connects securely to an API, specifies resource paths and filters, and manages data exchange formats and authentication.

To define an operation for the REST API tool, perform the following steps:

1Select the Parameters tab from within the REST API Tool connection.
2Click the Add Row icon and enter the operation details:

Property	Description
Key	Name of the query parameter. For example, consider the following API: https://api.openweathermap.org/data/2.5/weather?lat=xyz&lon=xyz&appid=xyz You can configure your parameters with the keys: lat, long, and appid.
Type	Type is always Query Parameter.
Data Type	Data type of the REST operation: - String - Integer - Float - Boolean
Description	Detailed description of the parameter.
Required	Whether the parameter is required or optional.

REST API Tool request headers

The REST API request header supplies the server with additional information like authentication credentials, content types, caching policies, and client preferences.

To define request headers for the REST API tool, perform the following steps:

1Select the Headers tab from within the REST API Tool connection.
2Click the Add Row icon and enter the header details:

Property	Description
Name	Name of the header, for example: Content-Type
Value	Value for the header, for example: application/json

REST API Tool request body

The REST API request body carries the data payload sent by the client to the server for creating, updating, or processing resources, formatted according to the specified content type.

You can define a default request body for the REST API Tool. However, this request body will be overridden in the following scenarios:

•If the REST API Tool has been added to the flow through a tool block, any variables defined for the request body are used instead of the default.
•If the REST API Tool has been added to an agent block, the request body is set to Agent Determines Value. In this scenario, the AI agent infers the request body from user input.
•If the user hasn’t provided any input for the request body and there are no variables declared for the request body, the agent uses the request body defined in the REST API Tool.

To define the request body for the REST API tool, perform the following steps:

1Select the Body tab from within the REST API Tool connection.
2Type or paste the body code into the editor. Alternatively, you can click the Upload icon to upload the request body from a file.

Authentication methods

The authentication method secures the API by confirming that requests come from valid and permitted clients, preventing unauthorized access.

To set an authentication method for the REST API tool, perform the following steps:

1Select the Authentication tab from within the REST API Tool connection.
2Select the Authentication Method to use:

Method	Description
Anonymous	Allows clients to access a REST API without authentication. This method is suitable for public or non-restricted resources. One example is a weather API that returns requests for current temperature data, without requiring a user token.
Basic	Uses a Base64-encoded username and password in the Authorization header. This method is not secure unless combined with HTTPS.
API Key	Uses a simple token provided by the API provider, which is passed in the request header or query parameters. This method is easy to implement but less secure if keys are exposed.

S3 File Parser and Loader tool

The S3 File Parser and Loader tool lets you access and ingest files from Amazon S3 into a data platform or application for further processing, analysis, or transformation.

The S3 File Parser and Loader tool connects to the specified S3 buckets using access credentials such as AWS Access Key and Secret Key. For a list of the supported file types that you can load, see S3 File Parser and Loader file formats.

You might use the S3 File Parser and Loader tool to accomplish the following tasks:

•Data ingestion for ETL (Extract, Transform, Load) processes
•Importing datasets for machine learning or analytics
•Migrating or archiving data from cloud storage

Examples: Real-time AI inferencing for dynamic document integration

You might use the S3 File Parser and Loader in the following cases:

•You are creating a customer service agent that needs to query policy documents in real-time.
•Your AI agent enables users to ask ad-hoc questions across a set of technical documentation.
•You are creating an interactive chatbot that requires instant document comprehension.

S3 File Parser and Loader file formats

The S3 File Parser and Loader tool is preconfigured to support several file formats. You can enable more formats with additional configuration.

The S3 File Parser and Loader supports the following file formats by default:

•CSV
•HTM
•HTML
•MD
•TXT

You can also use the following file formats with additional configuration:

•DOCX
•JPEG
•JPG
•PDF
•PNG
•PPTX
•XLSX

Enable additional file formats

You can enable additional file formats like CSV, DOCX, or JPEG to use with the S3 File Parser and Loader tool.

Perform the following tasks to enable additional file formats:

1Install Tesseract. This is an open source text recognition (OCR) engine that can extract text from images.
2Install the Unstructured library. This is an open source toolkit that can extract data from unstructured documents such as PDFs and Word files.

Step 1. Install Tesseract: You can install Tesseract on various operating systems. For example, to install Tesseract on Ubuntu/Debian, run the following commands:
Step 2. Install the Unstructured library: Perform the following steps:

File type	File extension
DOCX	docx
PDF	pdf
PPTX	pptx
XLSX	xlsx
PNG, JPG, JPEG	image

S3 File Parser and Loader connection details

You need to configure an S3 File Parser and Loader tool before you can use it in an AI agent.

The following table describes the properties of an S3 File Parser and Loader tool connection:

Property	Description
Name	Name of the connection to the S3 File Parser and Loader tool.
Location	Project or folder to save your assets. By default, assets are saved to the Default project.
Tool Type	Type of tool being configured.
Description	Optional. Description of the connection to the tool.
Bucket Name	Name of the specific S3 bucket containing the files you want to load.
Object Name	Specific name or key of the file (object) stored within an S3 bucket. The object name is also known as the key and can include folder-like paths, for example: data/2025/sales.csv The value for this field can be overridden by the AI agent when it invokes the tool. If you don't enter a value, the AI agent determines the object name when it invokes the tool.
AWS Access Key ID	Access key identifier associated with the AWS account or IAM user. This key works alongside the AWS Secret Access Key to authenticate and authorize requests. The AWS Access Key is a 20-character alphanumeric string, typically beginning with prefixes like AKIA or ASIA, for example: AKIAIOSFODNN7EXAMPLE
AWS Secret Access Key	Secret access key paired with the access key ID. The AWS Secret Access Key acts as a password paired with the Access Key ID. It is a long alphanumeric string typically around 40 characters in length, for example: wJalrXUtnFEMI/K7MDENG/bPxRfiCYEXAMPLEKEY

Salesforce Tool

The Salesforce Tool is used to query data in Salesforce.

Use the Salesforce Tool to perform queries on available data in Salesforce. For example, you need to fetch order details with a status of Draft.

Salesforce Tool connection details

You need to configure a Salesforce Tool before you can use it in an AI agent.

The following table describes the properties of a Salesforce Tool connection:

Property	Description
Name	Name of the connection to the Salesforce Tool.
Location	Project or folder to save your assets. By default, assets are saved to the Default project.
Tool Type	Type of tool being configured.
Description	Optional. Description of the connection to the tool.
Grant Type	OAuth 2.0 authentication method the client uses to request access tokens from an authorization server. You can select one of the following grant types: - Password. The client uses the user's username and password to request a token directly. This method is less secure and should only be used with trusted clients. - Client Credentials. The client authenticates directly with client ID and secret to obtain a token. Used for server-to-server communication with no user involvement.
Username	For Grant Types of Password, this is the Salesforce account username.
Password	For Grant Types of Password, this is the password for the Salesforce account.
Organization URL	For Grant Types of Client Credentials, this is the base URL of the Salesforce organization to which the tool connects to access resources. The following are examples of typical URLs: - For production or developer instances: https://api.salesforce.com - For sandbox or testing instances: https://test.salesforce.com - If your Salesforce organization uses a custom domain, it could resemble the following URL: https://yourcompany.my.salesforce.com
Client ID	Identifier for the connected app in Salesforce, used for OAuth authentication.
Client Secret	Secret key paired with the Client ID to authenticate the connected app. The Client Secret is typically a long, random string of characters generated when registering the client app with the authorization server.

Serper Search tool

The Serper Search tool is used to automate web searches, enabling data retrieval, research, or information gathering through APIs or integrations.

Use the Serper Search tool to search the web without manual browsing. This tool accepts search queries and returns structured search result data including links, snippets, titles, and metadata. It is often used to connect with external search providers or web scraping services to fetch live search results.

You might use the Serper Search tool to accomplish the following tasks:

•Enhancing chatbots or virtual assistants with the ability to fetch real-time web information.
•Building research tools that compile and analyze search results.
•Integrating search functionality within applications for improved user experience.

Example: Enhancing a virtual assistant

Your company's customer support AI agent integrates the Serper Search Tool to answer user queries with up-to-date web information. When a user asks a question, the AI agent sends the query through the Serper Search Tool, retrieves the most relevant search results, and uses the snippets to generate informative responses.

Serper Search connection details

You need to configure a Serper Search tool before you can use it in an AI agent.

The following table describes the properties of a Serper Search tool connection:

Property	Description
Name	Name of the connection to the Serper Search tool.
Location	Project or folder to save your assets. By default, assets are saved to the Default project.
Tool Type	Type of tool being configured.
Description	Optional. Description of the connection to the tool.
API Key	Unique key issued by Serper when you register for their API service. This key is used to authenticate your requests and track usage. The API Key is typically a long alphanumeric string that could contain special characters.

Spider Web Tool

The Spider Web Tool crawls and extracts data from websites, simulating how search engine spiders navigate the web.

The Spider Web Tool navigates through web pages by following links, discovering new pages and resources within the target URL, extracting specific elements such as text, images, metadata, or structured information like tables and lists.

This tool respects site crawling rules in a robots.txt file. Extracted data can be formatted, stored, or sent to other systems for further processing, analytics, or indexing.

The Spider Web Tool is preconfigured to navigate 25 pages beyond the base URL.

Example: Market intelligence collection

Your AI agent uses the Spider Web Tool to monitor competitor websites regularly. It crawls product pages to collect updated pricing, feature lists, and customer reviews. This data feeds into internal dashboards for competitive analysis and strategic decision-making.

Example: Content aggregation

Your AI agent uses the Spider Web Tool to crawl multiple news sites, extracting headlines and summaries to display on your platform, thereby keeping your content fresh and comprehensive.

Spider Web Tool connection details

You need to configure a Spider Web Tool before you can use it in an AI agent.

The following table describes the properties of a Spider Web Tool connection:

Property	Description
Name	Name of the connection to the Spider Web Tool.
Location	Project or folder to save your assets. By default, assets are saved to the Default project.
Tool Type	Type of tool being configured.
Description	Optional. Description of the connection to the tool.
API Key	Unique key that authenticates your Spider Web Tool connection and authorizes your access. It also identifies your account for usage tracking, billing, and error handling purposes. The API Key is typically a long alphanumeric string that could contain special characters.
Target URL	One or more URLs from which the spider begins crawling. Separate multiple URLs with a comma.
Limit	Maximum number of pages to crawl per website. Applies when the Mode is set to Crawl. If you set both the Limit and Depth thresholds, the web scraper stops as soon as either threshold has been reached. For example, if you set the limit to 1000 and the depth to 3, the web scraper stops when it has crawled either 1000 pages or has finished crawling all pages at a depth of 3, whichever comes first. Must be empty, 0, or a positive integer. If you don't specify a value or you set it to 0, no limit applies, so the tool crawls all pages.
Depth	Maximum number of link following levels the web scraper traverses from the starting URL. Applies when the Mode is set to Crawl. If you set both the Limit and Depth thresholds, the web scraper stops as soon as either threshold has been reached. For example, if you set the limit to 1000 and the depth to 3, the web scraper stops when it has crawled either 1000 pages or has finished crawling all pages at a depth of 3, whichever comes first. Must be empty, 0, or a positive integer. If you don't specify a value, the tool crawls 25 pages beyond the base URL. If you set this value to 0, no depth threshold applies.
Mode	Defines how the tool behaves on the target URL: - Scrape. The tool extracts information from the specified web page only. Doesn't follow links to other pages. This mode is best for quick data extraction or when you know exactly which pages to target. Example use case: Extracting product details from a specific e-commerce product page or gathering contact information from a directory page. - Crawl. The tool starts with the target URL and systematically follows links to discover and process multiple web pages. This mode is suitable for comprehensive site-wide data collection, indexing, or monitoring. Example use case: Crawling an entire news website to extract all articles published within the last month.

Creating a tool connection

Create tool connections as required for your AI agent.

1From the navigation menu, click New.

2In the New Asset dialog box, select Tool from the New Asset list.

3Select the tool to configure and then click Create.

4Enter the connection details.

Tip: Enter a name and description that will help you identify the tool when you select it within an AI agent or from a list. When you select a tool, the selection list displays the tool type as "Tool Connection," so it can be difficult to determine the tool type if the name or description doesn't identify the type.

5Click Save.