Last updated 13 October 2026
Use the help command to see Claude Code options
Accessing the internal command list
You are staring at a blinking cursor in your terminal, trying to remember the specific command to review your recent changes or adjust how much the agent can edit your files. It is frustrating to break your focus and search the internet for a list of flags and slash commands that should be right in front of you. When you are deep in a development flow, every second spent looking for documentation is a second where your mental model of the code begins to fade. You need a way to navigate the internal documentation of the Claude CLI to find hidden flags and configuration options without leaving your workspace.
The most direct way to see your options is to type /help inside an active session. This slash command displays a comprehensive list of all commands the agent understands in its current context. It covers everything from basic session management to advanced configuration. Instead of guessing whether a command exists, you can see the full syntax and a brief description of what each command does. This internal reference is the best source of truth because it matches the specific version of Claude Code you have installed on your machine.
Using /help frequently helps you discover features you might have missed during your first few sessions. For example, you can see how to use /review to have the agent look over your recent work or /model to check which version of the language model is currently driving the conversation. This command acts as a persistent map for the tool, ensuring you never have to leave the terminal to find the right syntax for your next task. It is the primary tool for any developer who wants to master the agent command line interface.
Configuring agent behavior and settings
Customizing how the agent interacts with your system is vital for maintaining a productive environment. You can use the /config command to view and modify the settings that govern the behavior of the agent. This command reveals how the agent handles file operations and which features are currently enabled. If you need to see the current state of your project or the overall health of the environment, the /status and /doctor commands provide immediate feedback. These tools allow you to verify that everything is running correctly before you start a complex coding task.
Your settings are stored in specific JSON files that the agent reads every time you start a session. A file named settings.json inside a .claude folder in your project root holds the configuration for that specific repository. If you have personal preferences that you do not want to commit to version control, you can place them in a settings.local.json file in the same directory. For settings that should apply to every project on your computer, the agent looks at ~/.claude/settings.json in your home folder. This hierarchy allows you to maintain strict project rules while keeping your personal workflow consistent across different codebases.
Managing project instructions with markdown files
The agent needs to understand the specific standards and architectural patterns of your project to be effective. You provide this context through a file named CLAUDE.md which sits in the root of your repository. This file serves as a manual for the agent, containing your preferred build commands, testing strategies, and code style requirements. When you start a new project, running the /init command helps you set up the initial structure and creates the necessary configuration files to guide the agent from the beginning.
You can refine these instructions by placing additional CLAUDE.md files in subdirectories. The agent prioritizes the instructions in the directory where it is currently working, allowing you to define different rules for a frontend folder versus a backend folder. If you have general rules that apply to every project you ever touch, you can create a CLAUDE.md file in the .claude directory of your home folder. This global instruction file is perfect for setting your default preferences for variable naming or documentation styles that you use across all your development work.
Controlling permissions and safety levels
Safety is a primary concern when giving an agent the ability to execute commands and write files on your machine. You can control these capabilities by switching between different permission modes. The default mode requires you to approve most actions, while the acceptEdits mode allows the agent to modify files more freely. If you want the agent to only suggest changes without executing them, you can use the plan mode. For environments where you want maximum speed and trust the agent completely, the bypassPermissions mode removes most interactive prompts.
You can cycle through these permission modes at any time by pressing Shift+Tab while in a session. This allows you to quickly tighten security when working on sensitive files or loosen it when performing repetitive refactoring tasks. You can also define specific permissions for each tool in your settings.json file. For example, you might choose to always allow the Read and Glob tools while requiring manual approval for the Bash and Write tools. This granular control ensures that the agent only does what you are comfortable with at any given moment.
Resuming previous work and history
One of the most useful features of the command line interface is the ability to pick up exactly where you left off. If you want to continue the very last session in your current folder, you can run the claude command with the --continue flag or simply -c. This saves you from having to re-explain the context of your task to the agent. If you need to find an older session, the --resume flag or -r flag provides a list of previous conversations in that directory. You can scroll through this list to find the specific task you were working on earlier in the week.
Inside an active session, the /resume command allows you to switch between historical sessions without restarting the tool. This is helpful when you realize that the answer to your current problem lies in a conversation you had two days ago. If you ever find that the agent is going down the wrong path, you can press Escape twice to see a list of earlier messages. Selecting an earlier message allows you to rewind the conversation to that point and try a different approach. This prevents you from getting stuck in a cycle of bad suggestions and helps maintain a clean project history.
Scaling multiple sessions with a visual wall
As your projects grow in complexity, you may find yourself managing several different features or bug fixes at the same time. Navigating these tasks through a single terminal window can become difficult to track. This is where adhdTerminal provides a visual solution for developers. It is a native app for macOS and Windows, made by MINDBALANCER OU in Estonia, not affiliated with Anthropic. It runs many Claude Code sessions side by side, where each one appears as a tile on a wall. Each tile has a color, a number, and a name, which helps you distinguish between different branches of your work at a glance.
The adhdTerminal app shows you the current status of every session, indicating whether the agent is working, waiting for your input, done, or idle. You can see the last few exchanges of each conversation directly on the tile, which makes it much easier to jump back into a task after a break. These tiles are persistent, meaning they reopen in the same place and resume the same session even after you restart your computer. It costs 99 dollars once, and one key runs one computer at a time and can be moved. There is a 7 day free trial and no account is required to start using it.
Monitoring resource usage and costs
Developing with an agent involves costs associated with token usage and processing time. You can keep a close eye on your spending by using the /cost and /usage commands during your session. These commands provide a detailed breakdown of how many tokens you have consumed and the resulting financial cost. Monitoring this data helps you understand which tasks are the most resource intensive and allows you to adjust your strategy if you are nearing a budget limit. It is a vital habit for developers who use the tool extensively for large scale projects.
The agent also provides a /compact command to help manage the size of your conversation context. As a session grows longer, the amount of data the agent must process increases, which can lead to higher costs and slower response times. Running /compact summarizes the conversation so far and removes unnecessary detail, freeing up space in the context window. This process can also be automated through the PreCompact hook event. By regularly compacting your sessions, you ensure that the agent remains fast and cost effective even during long development cycles.
Maintenance and mobile connectivity
Keeping your tools up to date ensures you have access to the latest improvements and security fixes. You can run the claude update command to check for the newest version of the CLI and install it automatically. If you want to see what has changed recently, the /release-notes command displays the latest updates directly in your terminal. For general troubleshooting, the claude doctor command checks your installation and environment to ensure everything is configured correctly for the agent to use tools like Bash, Read, and Write.
For developers who need to stay connected to their work while away from their desk, adhdTerminal offers a free iPhone app. This mobile companion mirrors your session wall over an end to end encrypted link, allowing you to monitor long running tasks from your phone. You can use the mobile app to answer a prompt from the agent or send the next task to a specific tile on your wall. This connectivity ensures that your development process can continue even when you are not physically at your computer, providing a bridge between your desktop environment and your mobile devices.
Questions people ask
How can I stop the agent if it starts making a mistake?
You can press the Escape key once to immediately interrupt the agent while it is working. If you need to revert the conversation to an earlier state, pressing the Escape key twice will bring up a list of previous messages. You can select a message from the past to jump back to that point and start over with a new prompt.
What is the fastest way to clear the current session context?
You can type the /clear command inside any active session to wipe the current conversation history. This allows you to start fresh in the same directory without closing the interactive interface. It is useful when you have finished one task and want to move on to another within the same project folder without carrying over old context.
How do I see the specific tools the agent is using?
The /help command provides a list of tools and slash commands available in your session. You can also use the /status command to see information about your current environment and the state of the agent. If you want to see the definitions for custom tools, you can check the .claude/agents or .claude/commands directories in your project.
Read next
- Add build instructions to your CLAUDE.md fileImprove project context by adding specific build commands and environment requirements to your CLAUDE.md file for the AI to follow.
- Add common error fixes to CLAUDE.mdImprove project memory by adding a section for common errors and their fixes to your CLAUDE.md configuration file.
- Add custom shell aliases for Claude CodeSpeed up your workflow by creating shell aliases for common Claude Code invocations and project-specific startup commands.
- Add test commands to a CLAUDE.md fileUnderstand how to define specific test runners and environment flags in your CLAUDE.md file so the tool runs the correct test suite.