Skip to main content
Superset is built for running multiple coding agents simultaneously. Each workspace provides a full-featured terminal environment where you can run any CLI-based agent.

​
Supported Agents

Superset works with any CLI-based coding agent. If it runs in a terminal, it runs in Superset.
AgentCommand Example
Claude Codeclaude-code
OpenCodeopencode
Cursor Agentcursor-agent
GitHub Copilotgh copilot
Gemini CLIgemini
Codex CLIcodex

​
Starting an Agent

1

Create or Switch to Workspace

Use ⌘N to create a new workspace or ⌘1-9 to switch to an existing one.
2

Open a Terminal

Press ⌘T to open a new terminal tab if needed.
3

Run Your Agent

Type your agent’s command and press Enter:
opencode "Implement user authentication"
4

Monitor Progress

Watch the terminal output as the agent works. Superset preserves all output history so you can scroll back anytime.

​
Terminal Management

​
Creating Terminals

ShortcutAction
⌘TNew terminal tab
⌘DSplit terminal right
⌘⇧DSplit terminal down
Use splits to monitor multiple agents side-by-side. For example, run a build watcher in one pane and an agent in another.
ShortcutAction
⌘⌥←Previous tab
⌘⌥→Next tab
⌘WClose current pane/terminal

​
Terminal Utilities

ShortcutAction
⌘KClear terminal
⌘FFind in terminal
All standard terminal features work: copy/paste, scrollback, ANSI colors, cursor positioning, etc.

​
Using Presets

Presets allow you to quickly launch common agent commands with a single keystroke.

​
Creating Presets

1

Open Preset Settings

Go to Settings > Terminal Presets or press ⌘/.
2

Add a Preset

Click Add Preset and configure:
  • Name: Display name (e.g., “Fix Bugs”)
  • Command: Shell command to run (e.g., opencode "Fix all TypeScript errors")
  • Shortcut: Ctrl+1 through Ctrl+9
3

Save and Test

Save the preset and press the shortcut to test it.

​
Using Presets

Press Ctrl+1 through Ctrl+9 to instantly run your configured presets in the current terminal.
Common preset ideas:
  • Ctrl+1: Run tests
  • Ctrl+2: Start dev server
  • Ctrl+3: Fix linting errors
  • Ctrl+4: Update dependencies

​
Running Multiple Agents in Parallel

Superset’s key strength is running multiple agents simultaneously without context switching.

​
Parallel Workflow

1

Create Multiple Workspaces

Press ⌘⇧N multiple times to quickly create workspaces for different tasks:
  • fix-auth-bug
  • add-dark-mode
  • update-dependencies
2

Start Agents in Each Workspace

Switch between workspaces with ⌘1-9 and start an agent in each:
# Workspace 1
opencode "Fix authentication redirect bug"

# Workspace 2
opencode "Implement dark mode toggle"

# Workspace 3
opencode "Update all dependencies to latest"
3

Monitor All Agents

The workspace sidebar shows which agents are:
  • Active (running)
  • Waiting (needs input)
  • Done (completed)
4

Review Changes as They Complete

When an agent finishes, press ⌘L to open the changes panel and review diffs.
Be mindful of system resources. Running 10+ agents simultaneously can be CPU/memory intensive depending on your machine and the agents you’re using.

​
Monitoring Agent Progress

Superset provides visual indicators for workspace state:

​
Status Indicators

IndicatorMeaning
🟢 Green dotAgent actively running
🟡 Yellow dotAgent waiting for input
⚪ Gray dotWorkspace idle
📝 Changes badgeUncommitted changes present

​
Changes Panel

Press ⌘L to open the changes panel, which shows:
  • Modified files: Files changed by the agent
  • Staged changes: Changes ready to commit
  • Diff viewer: Inline diffs for all changes
You can stage, unstage, and commit changes directly from the changes panel without leaving Superset.

​
Reviewing and Approving Changes

1

Wait for Agent to Complete

Watch the terminal output. Most agents will indicate when they’re done.
2

Open Changes Panel

Press ⌘L to view all changes made by the agent.
3

Review Each File

Click files in the changes panel to view diffs. Look for:
  • Correct implementation
  • No unintended changes
  • Proper code style
  • No security issues
4

Test Changes

Run tests or manually test the changes:
bun test
bun run dev
5

Commit or Request Changes

If satisfied, commit the changes:
git add .
git commit -m "Implement feature X"
Or ask the agent to make corrections:
opencode "Fix the validation logic in auth.ts"

​
Best Practices

One Task Per Workspace

Keep each workspace focused on a single task. This makes it easier to review changes and prevents agents from interfering with each other.

Clear Task Instructions

Give agents specific, clear instructions. Instead of “fix bugs”, say “fix the authentication redirect loop when logging out”.

Monitor Resource Usage

Keep an eye on CPU and memory usage. If your system slows down, consider running fewer agents in parallel.

Review Before Committing

Always review agent-generated code before committing. Agents can make mistakes or misunderstand requirements.

Use Presets for Common Tasks

Set up presets for frequently-run commands. This saves time and reduces typing errors.

Split Terminals for Monitoring

Use ⌘D and ⌘⇧D to split terminals. Run agents in one pane while monitoring logs or tests in another.

​
Troubleshooting

If an agent appears frozen:
  1. Press Ctrl+C to interrupt the process
  2. Check if the agent is waiting for input
  3. Look for error messages in the terminal
  4. Try running the agent with verbose logging
If the agent misunderstood your instructions:
  1. Review the changes with ⌘L
  2. Revert unwanted changes with git checkout -- <file>
  3. Give more specific instructions
  4. Break the task into smaller steps
If terminal output looks corrupted:
  1. Press ⌘K to clear the terminal
  2. Type reset and press Enter
  3. Restart the agent
If the agent is running but you don’t see output:
  1. Make sure you’re in the correct terminal tab
  2. Scroll to the bottom of the terminal
  3. Check if the agent is running in the background

Workspace Management

Learn how to organize multiple workspaces

Keyboard Shortcuts

Master all terminal shortcuts

Setup Scripts

Automate environment setup for agents

MCP Servers

Give agents access to external tools via MCP

Build docs developers (and LLMs) love

Get started for freeTalk to us