Run multiple Claude Code agents in parallel. Monitor status, track changes, and switch between codebases without losing context.
What It Does
Agent Fleet lets you orchestrate multiple Claude Code sessions from a single VS Code window. Each agent runs in its own terminal, working on a specific directory, while you monitor their status and changes in real time.
Key capabilities:
Multi-agent management: Add agents for different directories, repos, or git worktrees
Real-time status tracking: See when agents are idle, running, stuck (waiting for input), or complete
Git change detection: View uncommitted changes made by each agent directly in the tree view
Integrated terminals: Each agent gets a dedicated VS Code terminal running Claude Code
Workspace focus: Quickly switch your VS Code workspace to any agent's directory for full file exploration
Notifications: Get alerts when agents complete tasks or need attention
Who Should Use This
Agent Fleet is designed for developers who:
Work across multiple repositories or git worktrees simultaneously
Want to run parallel Claude Code sessions on different codebases
Need visibility into what multiple AI agents are doing without constantly switching terminals
Requirements
Claude Code CLI: Must be installed and available in your PATH (claude command)
VS Code 1.85.0 or higher
Getting Started
Install the extension
Click the Agent Fleet icon in the activity bar
Click "Add Agent" and select a directory
The extension will prompt you to install hooks on first use (required for status tracking)
Commands
Command
Description
Add Agent
Register a new directory as an agent
Remove Agent
Remove an agent from the fleet
Open Terminal
Open or focus the agent's Claude Code terminal
Focus Workspace
Add the agent's directory to your VS Code workspace
Install Hooks
Install Claude Code hooks for status tracking
Uninstall Hooks
Remove the installed hooks
How Status Tracking Works
Agent Fleet uses Claude Code's hook system to receive real-time status updates:
Status
Meaning
Idle
Terminal is open but not actively processing
Running
Agent is executing tools or generating responses
Stuck
Agent is waiting for user input (permission prompt, question, etc.)
Complete
Agent has finished its current task
Hooks are installed to ~/.claude/hooks.json and write events to a FIFO pipe that the extension monitors.