Tools menu
The Tools menu defines which tools the agent may use in the current chat and whether it must ask for confirmation before a call. Open it with the tools button next to the input field.
The menu settings apply to the agent's technical capabilities. A text request in a message can clarify the task, but it does not enable a disabled tool and does not override the confirmation set for it.
Tool categories
Tools are grouped by purpose:
| Category | What it is used for |
|---|---|
| Exploring Tools | Read files and directories, search for code, declarations and usages, study the project structure and dependencies |
| Editing Tools | Create and edit files, rename and move project elements |
| Terminal Tools | Run commands and manage processes in the terminal |
| Analysis Tools | Run code inspections and other analysis operations available in the IDE |
| Web Tools | Search the internet and read pages |
| Configuration tools | Find and run IDE configurations |
Allow Analysis Tools when the agent has to check code with the IDE's own means, for example before finishing an implementation or during a review. For minimal access, keep only the analysis operations you need and enable confirmation for actions that may execute code. For more about the available checks, see the tool catalog.
Allow Web Tools when a task needs up-to-date external information or a read of a known primary source. Keep only search or page fetching, depending on whether you know the URL. The differences between these scenarios are described on the Web search page.
Allow Configuration tools when the agent has to pick an existing IDE configuration and run a build, a test or an application with the configured SDK, classpath and environment variables. The minimal set is getting the list of configurations and running the selected one; keep confirmation on for the run if it can change files or external systems. For details, see Run configurations.
Below the built-in categories, the menu shows the connected MCP servers, for example chrome or searxng. Their tools are configured separately from the built-in ones. Use Add more tools when the built-in capabilities are not enough and the agent needs access to an external system, such as an issue tracker or a browser. Open Add more tools, connect the MCP server you need and allow only the tools of that server required for the current task. Configuring servers is described on the MCP page.
The set of categories depends on the Explyt version, the selected agent mode, the IDE and the connected MCP servers. If the tool you need is missing from the menu, check the feature matrix, the agent mode and the MCP settings.
Tool access mode
For each tool you can choose one of three options:
| Setting | What happens |
|---|---|
| Allowed without confirmation | The agent can call the tool and continue working without stopping |
| Ask for confirmation | Before the call, the agent stops and asks the user for permission |
| Forbidden | The agent cannot call the tool in this chat |
If confirmation is enabled for a tool, Explyt stops right before the call. Once the user answers, the agent either performs the action or continues without it, if the task allows another path.
Do not assume that every available tool is automatically safe. For example, a terminal command, a test or a run configuration can change files, a database or an external system. For potentially irreversible actions, leave confirmation enabled.
How to configure Tools before a task
- Open the Tools menu.
- Expand the category you need.
- Forbid the tools the task does not need.
- Enable a confirmation prompt for the actions you want to control.
- Leave calls without confirmation only for actions whose consequences you understand.
Examples:
- for explaining code, you can keep Exploring Tools and forbid Editing Tools and Terminal Tools;
- for a small edit, allow reading and editing but keep confirmation for the terminal and for runs;
- before working with a production system, forbid the terminal, the browser tools and any configurations that can reach it.
How Tools differs from other restrictions
The Tools menu is responsible for the type of action, while the other mechanisms cover files and instructions:
- Edit Scope restricts which files and directories can be changed in the current task;
.readignoreand.writeignorepermanently forbid reading or writing under the listed paths;- the agent mode sets the role and the baseline set of available tools;
- Rules, Skills and AGENTS.md give the agent instructions, but by themselves they do not enable or disable tools.
For a sensitive task, use the restrictions together: choose a suitable agent mode, configure Tools, set Edit Scope and block secrets with .readignore.
If the agent stops and asks for permission
Check:
- which tool the agent is about to call;
- which parameters or command are shown in the call card;
- which files, processes or external systems the action can change;
- whether this call is needed for your task.
Allow the action only after checking. If the same safe call keeps repeating and you do not want to confirm it every time, change that tool's setting in the Tools menu. For dangerous and irreversible actions, do not turn confirmation off for convenience.
For more about reading, writing, runs and reviewing changes, see the Permissions and safety page.