MultiTerminal
Agent terminal
Run your preferred AI coding CLI and keep its progress, questions, and proposed changes visible.
AI coding workflows
Keep your AI coding loop visible from prompt to review.
Run your coding agent beside the development server, tests, and running app. Follow a changing Markdown plan or release note without leaving the workspace.
The problem
Keep agent questions and test failures in view.
While an AI coding tool edits a feature, another terminal may be serving the app, a watcher may be running tests, and a browser may be showing the result. Separate windows make it easy to miss a question, an error, or a regression.
MultiTerminal gives each part of the loop a pane. You can watch the agent work, inspect runtime output, test the result in a real embedded browser, and follow a Markdown plan or release note as it refreshes.
Example layout
A four-pane AI coding workspace
Runtime terminal
Serve the app or API and watch requests, rebuilds, warnings, and failures as they happen.
Tests in a tab group
Group a test watcher with another terminal so validation is close without consuming a permanent split.
Browser and live notes
Open the localhost app beside a Markdown preview that refreshes when the agent or editor changes it.
Workflow
A reviewable loop, not a hidden one
- Start the agent, development server, and test watcher in terminals with the right project folders.
- Keep the running app open in a browser pane and reproduce the behavior the agent is changing.
- Watch tests and runtime output while the work proceeds; switch grouped tabs when a pane does not need constant attention.
- Review the app, file preview, and terminal results before accepting or committing changes.
Try this setup
Agent, runtime, tests, and release notes.
This example uses a fictional Launchpad project and separately installed tools. Use the folders and scripts defined in your own project.
Agent terminal
Set its starting folder to the project root. Run your installed agent CLI, such as Codex.
cd ~/Projects/launchpad
codexServer and test tabs
Keep the server visible and group a test watcher into a neighboring tab. These sample commands require matching package scripts.
npm run dev
npm run test:watchAdd a Browser pane for the address printed by your server. Add a File Preview pane for your Markdown plan or release notes. Review browser behavior, test output, and the agent’s proposed changes before committing.
