Last updated 13 October 2026
Managing multiple Claude Code projects
Context switching between multiple terminal windows
Working on multiple software projects at the same time is a common reality for developers. You might have a frontend repository, a backend service, and a shared library all open simultaneously. When you use Claude Code to assist with these projects, you often end up with several terminal tabs or windows. Each window represents a different session with its own specific history and context. This creates a mental burden because you must remember exactly what each agent is doing and what files it has already edited.
The primary challenge is context switching. Every time you move from one terminal tab to another, you have to spend a few seconds orienting yourself. You need to check the last few lines of the conversation to see if the agent completed the Task tool execution or if it is waiting for your input. If you are handling three or four sessions, it is easy to lose focus or make a mistake by typing a command into the wrong window. This friction reduces the efficiency you gain from using an AI coding agent.
Project organization with CLAUDE.md
The CLAUDE.md file is your primary tool for keeping different projects isolated and well defined. You should place this file in the root of every repository. It serves as the source of truth for the agent, containing build commands, testing procedures, and specific coding standards for that project. When you switch between a Python project and a TypeScript project, the agent reads the local CLAUDE.md to understand which Bash commands are appropriate for the environment.
If you have a large monorepo, you can place additional CLAUDE.md files in subdirectories. This allows you to give the agent more specific instructions for a particular part of the tree. For example, the instructions for a legacy service might differ significantly from a new greenfield project. The agent automatically respects these boundaries, which prevents it from applying the wrong patterns to your code. If you have preferences that should apply across all your work, you can also use a global file located at ~/.claude/CLAUDE.md.
Managing a wall of sessions in adhdTerminal
Managing multiple sessions through standard terminal tabs often leads to projects being forgotten in the background. You might not notice that an agent has finished its work or is stuck waiting for a permission to use the Write tool. adhdTerminal solves this by displaying every active Claude Code session as a tile on a visual wall. This native app for macOS and Windows runs many sessions side by side, allowing you to see the status of every project at a single glance.
Each tile on the wall can be assigned a name, a number, and a colour. This visual organization helps you quickly identify which project is which. The tiles show whether the agent is working, waiting for you, done, or idle. They also display the last few exchanges of the conversation, so you can see exactly what the agent is currently doing without having to click into the window. This reduces the cognitive load of tracking multiple concurrent streams of work.
Resuming and continuing work
When you move between projects, you do not always want to start a fresh conversation. You might need to pick up a task that you started yesterday or work on a different branch. To continue the most recent session in a folder, use the command claude --continue or the shorter version claude -c. This reloads the context and history so you can keep working without repeating your previous instructions. It is a vital command for maintaining continuity across a long development cycle.
If you need to access a specific historical session that is not the most recent one, use the command claude --resume or its short form -r. This provides a list of earlier sessions for you to choose from. This is particularly useful when you are juggling multiple features in the same repository. By resuming the correct session, you ensure the agent has the specific context of the feature you are working on, including any files it has already read using the Read or Glob tools.
Resource monitoring and context management
Running multiple sessions can quickly consume a lot of resources and increase your token spend. Each conversation fills the context window with the text of the chat and the contents of the files the agent has accessed. To keep track of this, use the /usage and /cost commands inside your sessions. These commands provide a clear picture of how many tokens you have used and the financial cost associated with the current project. Monitoring these figures helps you decide when it is time to reset a session.
When a session becomes too large and slow, the /compact command is your best tool. It summarizes the conversation so far and frees up space in the context window. This allows the agent to continue working effectively without losing the core progress it has made. If you want to start completely over in the same folder while keeping your project instructions, use the /clear command. This is helpful when you finish one large task and want to begin the next one with a clean slate.
Permission tiers and configuration files
Different projects often require different levels of trust. For a critical production database service, you might want to review every change, while for a personal prototype, you might want the agent to work faster. You can cycle through permission modes like default, acceptEdits, plan, and bypassPermissions by pressing Shift+Tab. These modes determine how often the agent asks for your approval before using tools like Edit or Bash to modify your files or run commands.
You can make these settings permanent by editing the .claude/settings.json file in your project directory. For example, you can specify which tools are allowed to run without a prompt. If you have personal overrides that you do not want to check into version control, use the .claude/settings.local.json file. Global settings that apply to every project on your computer should be placed in ~/.claude/settings.json. This tiered configuration system gives you granular control over how the agent behaves in different environments.
Tool integration with MCP
Managing multiple projects often requires access to specialized external tools. You can manage Model Context Protocol servers using the claude mcp command from your shell or the /mcp command within a session. These servers allow the agent to interact with external data sources or specialized utilities that are not part of the standard toolset. The configuration for these servers is stored in a file named .mcp.json.
By using project specific MCP configurations, you can ensure that the agent only has access to the tools it needs for the current task. For instance, a backend project might have an MCP server for querying a specific database, while a documentation project might have one for searching a knowledge base. This isolation improves security and prevents the agent from getting confused by too many available tools. You can use the /mcp command to check the status of these connections at any time.
Persistent tiles and mobile monitoring
If you are running many sessions, a system restart can be a major interruption. adhdTerminal handles this by ensuring that your project tiles reopen in the same place and resume the same session after a restart. This persistence means you do not have to manually set up your workspace every morning. The app, created by MINDBALANCER OU in Estonia, is designed for developers who need a stable and organized environment for their AI assisted workflows.
To stay informed while you are away from your desk, you can use the free iPhone app that mirrors your project wall. The link is end to end encrypted for security. From the mobile app, you can check the status of your running tasks and even answer a prompt if the agent is waiting for a permission. This is especially useful for long running tasks that use tools like WebSearch or Grep across large codebases. You can provide the next task or a quick clarification from anywhere.
Automating workflows with custom commands
You can save time across different repositories by creating custom slash commands. These are stored as markdown files in the .claude/commands directory within your project. A custom command can define a sequence of actions or a specific way you want the agent to handle a task. For example, you could create a command that runs a specific test suite and then uses the TodoWrite tool to record any failures it finds.
Standardizing these commands across your projects creates a consistent interface for your work. Instead of typing out long instructions every time, you can simply run your custom slash command. This is particularly effective for repetitive tasks like code reviews or documentation updates. Since these commands are just markdown files, they are easy to create, edit, and share across your different project repositories. This modular approach helps you maintain a high level of productivity regardless of which project you are focusing on.
Questions people ask
How can I stop an agent that is stuck in a loop?
You can press the Escape key once to interrupt what the agent is currently doing. If you need to exit the session entirely, use the Ctrl+C shortcut in your terminal. For deeper issues with the installation, you can run the claude doctor command to check the health of the setup.
What is the best way to move between past messages?
If you press the Escape key twice, the interface will offer you a list of earlier messages in the current session. You can then select a previous point in the conversation to go back to. This is helpful if the agent has taken a wrong turn and you want to restart from a known good state.
How do I make sure the agent ignores certain files in my project?
Claude Code respects your .gitignore file by default, so it will not read or modify files that you have excluded from version control. If you need to hide additional files specifically from the agent, you should ensure they are covered by your git ignore rules. The agent uses the Glob and Grep tools to find files, and these tools will skip ignored paths.
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.