Try free

Last updated 13 October 2026

See project headlines in the terminal

The mental cost of invisible context in terminal sessions

When you manage multiple coding tasks with Claude Code, your workspace quickly becomes a collection of identical terminal tabs. You might have one session refactoring a legacy module, another running a suite of unit tests, and a third searching for a bug in the configuration files. Each window shows a similar bash prompt and a stream of tool outputs. This lack of visual distinction forces you to spend mental energy every time you switch tasks. You have to scroll through the history to remember the last prompt or wait for the agent to finish its current loop before you can identify which session you are looking at.

For a developer with ADHD, this constant re-orientation is a major source of friction. The effort required to pick up the thread of a conversation can lead to distraction or procrastination. You might lose the momentum you had on a difficult feature just because you spent two minutes trying to figure out which terminal window was handling the database migration. Without a clear and immediate summary of what each agent is doing, your productivity drops as the number of parallel sessions increases.

Using project instructions for consistent status reporting

The most effective way to keep track of your work is to make the agent responsible for its own status updates. You can achieve this by adding specific rules to your CLAUDE.md file. This file sits in the root of your repository and provides persistent instructions that the agent reads at the start of every session. If you are working on a large project with multiple components, you can also place a CLAUDE.md file in a subdirectory to provide localized instructions for that specific part of the tree.

In your root CLAUDE.md, instruct the agent to maintain a summary of its current objective. You should tell the agent to use the Bash or Write tools to update a specific file with its progress. For example, you can require the agent to write a single line of text to a file named STATUS.md every time it completes a major step. This ensures that even if you walk away from your computer, there is a clear record of the last action taken and the next step in the plan. This protocol turns the agent into a self-reporting worker that manages its own visibility.

Defining a headline protocol for the agent

To make the status updates useful for a dashboard, you need to define a clear format for the headlines. Tell the agent to provide specific numbers and brief descriptions of its current state. If the agent is running tests using the Bash tool, it should report how many tests passed and how many failed. If it is performing a global search with the Grep tool, it should report how many matches it found and which file it is currently processing. These small pieces of data act as project headlines that give you an instant pulse on the work.

You can also instruct the agent to use the TodoWrite tool to keep a list of remaining tasks. By forcing the agent to update this list frequently, you create a living document of the session history. When you return to a session after an interruption, you can quickly read the latest entries in your status file to regain context. This approach moves the information from the agent's internal memory into the project files where it is easily accessible to you and other tools in your environment.

Automating status updates with hook events

Claude Code provides several hook events that allow you to run scripts at specific points in the conversation loop. You can use the PostToolUse or UserPromptSubmit hooks to trigger scripts that extract headlines from the agent's output. These hooks are configured in your .claude/settings.json file or your global ~/.claude/settings.json file. By setting up a script that runs after every tool execution, you can automatically parse the output for specific patterns and surface them to your desktop environment.

For instance, you can create a hook for the PostToolUse event that looks for strings like progress or status in the agent's response. The script can then format this information and write it to a standard location. This automation ensures that your project headlines are always up to date without you having to manually prompt the agent for a summary. It creates a feedback loop where the agent's work is constantly being summarized and presented in a way that is easy to digest at a glance.

Visualizing parallel sessions with adhdTerminal tiles

Standard terminal emulators are not designed to show you the state of many parallel processes at once. This is where adhdTerminal provides a specialized solution. It is a native app for macOS and Windows, created by MINDBALANCER OU, that runs many Claude Code sessions side by side. Instead of hidden tabs, every session is a tile on a wall. Each tile has a specific color, number, and name, making it impossible to confuse one task with another. You can see whether a session is working, waiting for your input, done, or idle without clicking into it.

The strength of the tile wall is its ability to display the numbers and headlines your project publishes. Because the tiles show the last exchanges of the conversation, you can see the latest status updates directly on the dashboard. If you have instructed the agent to report its progress through the Bash tool or Write tool, those headlines appear right under the tile for that project. This setup allows you to monitor a dozen different tasks simultaneously, moving your focus only when a tile indicates that it is waiting for your next command.

Managing session continuity and resumption

When you work on complex features, you often need to stop and restart your terminal. Claude Code handles this with the claude --continue and claude --resume commands. Using claude --continue carries on the most recent session in your current folder, while claude --resume gives you a list of previous sessions to pick from. This is useful for picking up where you left off, but it still requires you to remember which session was which based on a list of timestamps or IDs.

Within the tile-based interface, this process is simplified because tiles reopen in the same place after a restart and resume the same session automatically. You do not have to manually run resume commands for every project. The visual position of the tile becomes a mnemonic for the task. If the top-left tile is always your API debugging session, you will naturally look there for updates. This spatial consistency reduces the cognitive load of managing long-running projects and helps you maintain your focus over several days of work.

Monitoring tasks remotely with mobile integration

The ability to step away from your desk while the agent performs a long task is essential for maintaining a healthy work rhythm. The adhdTerminal system includes a free iPhone app that mirrors your wall of tiles over an end-to-end encrypted link. This allows you to check the status of your coding sessions from another room. If an agent finishes a task and publishes a headline saying that all tests have passed, you will see that update on your phone immediately.

The mobile app also allows you to interact with your sessions if the agent reaches a point where it needs your permission or a new task. You can answer a prompt or send the next command directly from the iPhone app. This means you do not have to stay tethered to your desk just to keep the agent moving. You can manage the high-level flow of your project headlines from anywhere, ensuring that your parallel sessions continue to make progress while you take a break or handle other responsibilities.

Optimizing context and cost during long sessions

As your sessions grow and the agent performs more work, the context window fills up with conversation history and file contents. This can make the agent slower and more expensive to run. You should use the /cost and /usage commands regularly to monitor your spend. If a session becomes too heavy, use the /compact command. This summarizes the conversation so far and frees up space in the context window, which helps the agent stay focused on the current task and reduces the tokens used for each new prompt.

Keeping your sessions lean also ensures that the project headlines the agent generates remain accurate. If the context is too cluttered, the agent might lose track of the specific reporting rules you set in CLAUDE.md. By using /compact and occasionally using /clear to start over in the same folder, you maintain a high level of performance. This discipline ensures that your dashboard remains a reliable source of truth for all your active developer tasks.

Questions people ask

How do I make the agent show specific numbers on my dashboard?

You should add instructions to your CLAUDE.md file telling the agent to use the Bash or Write tools to output status updates in a specific format. If the agent outputs a line starting with a keyword like Status or Progress, a dashboard like adhdTerminal can pick up that headline and display it on the project tile. This allows you to see real-time metrics like test pass rates or completion percentages without opening the terminal.

Can I use multiple Claude Code sessions in the same folder?

Yes, you can run multiple sessions in the same folder by starting new instances of the command line agent. However, to keep them organized, it is better to use a tile-based interface that assigns a unique name and color to each session. This prevents you from accidentally overwriting work or getting confused about which session is performing which specific sub-task in that directory.

What happens if I reach the context limit during a long task?

When the context window is nearly full, you should use the /compact slash command inside the session. The agent will then summarize the work it has done so far and remove the detailed history of previous exchanges to save space. This keeps the session responsive and allows the agent to continue following the instructions in your CLAUDE.md file without failing due to a full context window.

Read next

More on claude code basics·All posts