Run terminals and tasks

On this page

Keep agent work, a development server and test output accessible from the same workspace.

Before you start

Open the repository workspace. A Quest is a saved task from its .code-workspace configuration; Bash Terminal is an interactive shell.

Start a shell or task

  1. Open Summon in Command/Action mode.
  2. Select Bash Terminal for an interactive shell, or select a saved task under Quests.
  3. Choose the working directory when prompted.
  4. Run the project's documented command, such as its development server or test script.
  5. Launch another terminal or agent when a second process is needed. Each instance has its own output.

To save a repeatable command, add its entry under tasks in the .code-workspace file, as described in the workspace reference. You can ask an agent directly in its terminal to prepare the entries, preserving the existing folders and settings. Inspect the command and working directory, then reopen Summon to launch the saved Quest.

The recording uses these two entries in a workspace with a folder named app:

"tasks": {
  "version": "2.0.0",
  "tasks": [
    {
      "label": "Test Ember app",
      "command": "pwd && npm test",
      "options": { "cwd": "${workspaceFolder:app}" }
    },
    {
      "label": "Serve Ember app",
      "command": "python3 -u -m http.server 4181 --bind 127.0.0.1",
      "options": { "cwd": "${workspaceFolder:app}" }
    }
  ]
}

The server command serves the static demonstration app. Use your own project's development command and an available port. The test Quest, server Quest and interactive shell each keep separate output. Stop the server with Ctrl+C in its terminal when finished.

Focus a terminal and press Ctrl/Cmd+F to search its buffer. Select a failure and copy it with Ctrl/Cmd+Shift+C, or send the selection to another session using Ctrl/Cmd+Shift+1…9. Open the Spellbook with Ctrl/Cmd+Shift+/ to check the shortcuts available on your platform.

Example instruction

Inspect this test failure, identify the cause in the current repository and propose the smallest fix. After editing, rerun the failing test and report its result.

Arrange your work

Use sidebar agent groups, session tabs and separate workspace windows. Drag tabs to reorder them. Ask an agent to show the relevant Git diff in its terminal when comparing code changes.

The + and caret next to the tabs form a two-part add/resume button. They do not split the terminal area. Arbitrary layouts mixing terminals, browsers and editors are not documented as an equivalent to Orca's split-anything feature.

Check the result

Switch between sessions and confirm their processes and output remain distinct. If a command ran in the wrong folder, stop that task and relaunch it with the correct working directory.

More detail

Agents and Tasks · Toolbar and shortcuts