.png?fit=max&auto=format&n=9PuWEwunOBJ8KxNm&q=85&s=44640a56acb168b190ee551b29b33392)
To make a component into a tool that an agent can use, enable **Tool Mode** in the component. Enabling tool mode modifies a component input to accept calls from an agent. If the component you want to connect to an agent doesn't have a **Tool Mode** option, you can modify the component's inputs to become a tool.
## **arXiv**
This component searches and retrieves papers from [arXiv.org](http://arXiv.org).
The **Tool Name** and **Tool Description** fields are required for the Agent to decide when to use the tool. **Tool Name** cannot contain spaces.
The values for **Collection Name**, **Astra DB Application Token**, and **Astra DB API Endpoint** are found in your Astra DB deployment. For more information, see the [DataStax documentation](https://docs.datastax.com/en/astra-db-serverless/databases/create-database.html).
In this example, an **OpenAI** embeddings component is connected to use the Astra DB tool component's **Semantic Search** capability. To use **Semantic Search**, you must have an embedding model or Astra DB Vectorize enabled. If you try to run the flow without an embedding model, you will get an error.
Open the **Playground** and ask a question about your data. The Agent uses the **Astra DB Tool** to return information about your collection.
### **Define Astra DB tool parameters**
The **Tool Parameters** configuration pane allows you to define parameters for [filter conditions](https://docs.datastax.com/en/astra-db-serverless/api-reference/document-methods/find-many.html#parameters) for the component's **Find** command.
These filters become available as parameters that the LLM can use when calling the tool, with a better understanding of each parameter provided by the **Description** field.
1. To define a parameter for your query, in the **Tool Parameters** pane, click **Add a new row**.
2. Complete the fields based on your data. For example, with this filter, the LLM can filter by unique `customer_id` values.
* Name: `customer_id`
* Attribute Name: Leave empty if the attribute matches the field name in the database.
* Description: `"The unique identifier of the customer to filter by"`.
* Is Metadata: `False` unless the value is stored in the metadata field.
* Is Mandatory: `True` to require this filter.
* Is Timestamp: `False` since the value is an ID, not a timestamp.
* Operator: `$eq` to look for an exact match.
If you want to apply filters regardless of the LLM's input, use the **Static Filters** option, which is available in the component's **Controls** pane.
| **Parameter** | **Description** |
| :------------- | :------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
| Name | The name of the parameter that is exposed to the LLM. It can be the same as the underlying field name or a more descriptive label. The LLM uses this name, along with the description, to infer what value to provide during execution. |
| Attribute Name | When the parameter name shown to the LLM differs from the actual field or property in the database, use this setting to map the user-facing name to the correct attribute. For example, to apply a range filter to the timestamp field, define two separate parameters, such as `start_date` and `end_date`, that both reference the same timestamp attribute. |
| Description | Provides instructions to the LLM on how the parameter should be used. Clear and specific guidance helps the LLM provide valid input. For example, if a field such as `specialty` is stored in lowercase, the description should indicate that the input must be lowercase. |
| Is Metadata | When loading data using LangChain or LLM Controls, additional attributes may be stored under a metadata object. If the target attribute is stored this way, enable this option. It adjusts the query by generating a filter in the format: `{"metadata.| Name | Display Name | Info |
| input\_value | Input | |
| lang | Language | |
| k | Number of results | |
| load\_all\_available\_meta | Load all available meta | |
| doc\_content\_chars\_max | Document content characters max |
| Name | Display Name | Info |
| input\_value | Input Query | Example query: 'What is the population of France?' |
| app\_id | App ID |