Terminal Setup & Monitoring
Multi-Tab Terminal Workflow
The aitasks framework is built for terminal-centric development. Using a terminal emulator that supports multiple tabs or panes — switchable with keyboard shortcuts — makes the workflow significantly more efficient.
Some good terminal emulators:
- Warp — Modern terminal with built-in Claude Code integration, multi-tab support, and real-time diff viewing. Available for Linux, macOS, and Windows
- tmux — Terminal multiplexer with split panes and sessions. Works everywhere
- Ghostty — Fast GPU-accelerated terminal with tabs and splits
- WezTerm — Highly configurable GPU-accelerated terminal with built-in multiplexer, tabs, and panes. Available for Linux, macOS, and Windows
Typical tab layout:
| Tab | Purpose |
|---|---|
| Tab 1 | Main code agent session running /aitask-pick |
| Tab 2 | ait board for visual task management and triage |
| Tab 3 | ait create ready to launch for capturing new ideas |
| Tab 4 | Git status / diff viewer (lazygit, delta, or plain git diff) for monitoring implementation changes |
IDE alternative: You can also run a terminal inside your IDE (VS Code, IntelliJ, etc.) and use another pane to watch file changes in real time. However, dedicated terminal emulators with keyboard-driven tab switching tend to be faster for this workflow.
Monitoring While Implementing
While /aitask-pick is running — especially during the exploration or implementation phases which can take several minutes — you can stay productive in other terminal tabs.
What to do while waiting:
- Triage tasks — Open
ait boardin another tab to review priorities, move tasks between kanban columns, update metadata (priority, effort, labels), and adjust dependencies. See the Board documentation for all available operations and keyboard shortcuts - Capture new ideas — As ideas come up during the implementation (which they often do while watching the agent work), quickly switch to a tab with
ait createand write them down. The key shortcutninait boardalso launches task creation directly - Review progress — Watch the current diff in another tab to understand what changes are being made. lazygit, Warp’s built-in diff viewer, or a plain
git diffall work well for this
This parallel workflow means the human never becomes a bottleneck waiting for the AI agent to finish. You are always either reviewing the agent’s output, managing your task backlog, or capturing the next set of ideas.
Context Monitoring
One of the key advantages of decomposing work into small connected tasks is reduced context usage — whichever code agent you use will generally perform better when it has more room in its context window. Monitoring context usage in real time helps you understand when a task is getting too large and should be split.
Recommended plugin:
- claude-hud — Claude Code plugin that displays real-time context window usage directly in your terminal. Useful if Claude Code is your primary agent. Shows token count, percentage filled, and alerts when context is getting large