> ## Documentation Index
> Fetch the complete documentation index at: https://docs.lighton.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Tool Management

> This page allows System Administrators to configure the agent’s functional capabilities at the global instance level. Tools are the functions the AI model can call to answer user requests (e.g., searching the web, analyzing a document, executing code).   This section focuses on Native Tools, developed by LightOn and integrated by default into Paradigm.

## **1. Overview and Tool List**

The main **“Tools”** page displays the full inventory of available capabilities on your instance.

### **List organization**

Tools are grouped by origin:

* **Global / Native Tools:** The 8 core tools provided by LightOn (e.g., `code_execution`, `web_search`, `document_search`)
* **MCP Tools:** Tools connected via external **MCP (Model Context Protocol)** servers (e.g., Zendesk connectors, Mail, etc.)

### **Quick indicators**

For each tool, two key columns indicate its status:

* **Included in routing:** (Green check) Indicates whether the agent is allowed to choose this tool on its own
* **User selectable:** (Green check) Indicates whether the end user can manually force the use of this tool in the chat interface

### **Filters**

A right-side panel allows you to filter the list by name, by company, or by properties (e.g., show only tools marked as **“Included in routing”**).

## **2. Configuring a Native Tool**

By clicking on a tool name (e.g., `document_search`), you access its detailed configuration. These settings define how and when the tool can be used.

### **A. Basic settings and constraints**

* **Require document:** Defines whether the tool requires a document to be uploaded by the user in order to work (e.g., `True` for document analysis)
* **Accepted file types:** Lists the file extensions supported by this tool (e.g., `pdf`, `ppt`, `docx`, `md`, `jp`)

### **B. Activation rules (Routing & Selection)**

Three checkboxes determine the tool’s accessibility:

* **Included in routing:** If checked, the model (“the brain”) can autonomously decide to use this tool to answer a question
* **User selectable:** If checked, the tool will appear in the user interface, allowing the user to explicitly select it
* **Enabled by default:** If checked, the tool will be enabled by default in new chat settings (**Chat Settings**)

### **C. Sequencing policy (Policy)**

This advanced section controls tool chaining (“What is the agent allowed to do after using this tool?”):

* **Allowed Tools:** Whitelist of tools allowed to run after this one
* **Disallowed Tools:** Blacklist of tools forbidden after this one
* **Should Exit Reasoning:** Forces the agent to stop reasoning and generate a final response after running this tool

## **3. List of Standard Native Tools**

Here are the main native tools you can administer:

| Tool                         | Description                                                                             |
| :--------------------------- | :-------------------------------------------------------------------------------------- |
| **web\_search**              | Allows the agent to search for real-time information on the Internet.                   |
| **document\_search**         | Semantic search within documents uploaded by the user.                                  |
| **document\_analysis**       | In-depth analysis and full reading of files.                                            |
| **code\_execution**          | Generates and executes code (often Python) for complex calculations or data processing. |
| **adaptive\_clarification**  | Allows the agent to ask the user questions to clarify an unclear request.               |
| **in\_context**              | Uses the immediate context of the conversation.                                         |
| **vision\_document\_search** | Specialized search for documents containing images or visual elements.                  |

### **Admin note**

Changes made here (especially **“Enabled by default”**) affect the baseline configuration of all new conversations, or the default **Chat Settings** for companies within the instance.