TaskSync
Queue your prompts. Human in the loop workflow
TaskSync lets you batch and queue your prompts to AI agents in VS Code, so they can keep working while you stay focused. Perfect for long-running tasks, repetitive workflows, or hands-free automation—saving you time and reducing premium requests.
Features
Smart Queue Mode
Queue multiple prompts to be automatically sent when the AI agent requests feedback. Perfect for:
- Batching instructions for long-running tasks
- Pre-loading prompts for predictable workflows
- Reducing interruptions during focused work
Normal Mode
Direct interaction with AI agents - respond to each request as it comes in with full control over the conversation flow.
Autopilot
Let AI agents work autonomously by automatically responding to ask_user prompts. When enabled:
- Add multiple prompts that cycle in order (1→2→3→1...) with each
ask_user call
- Drag to reorder, edit, or delete individual prompts in the Settings modal
- Toggle on/off from the Autopilot switch below the send button, or in Settings
- Queue priority: queued prompts are ALWAYS sent first — Autopilot only triggers when the queue is empty
- Perfect for varying instructions mid-session or hands-free operation on well-defined tasks
Response Timeout (Auto-respond when idle)
Prevent tool calls from waiting indefinitely when you're away:
- Configure timeout duration in VS Code Settings, including disabled (
0 minutes), 5 minutes, and options up to 240 minutes (4 hours)
- When timeout elapses, TaskSync auto-responds with Autopilot text
- Consecutive limit: After N consecutive immediate Autopilot responses (configurable, default 5), Autopilot is automatically disabled to prevent infinite loops
- Timeout-based auto-responses do not count toward this consecutive limit
- Counter resets when you manually respond
Human-Like Delay
Simulate natural human pacing by adding random delays before automated responses:
- Enabled by default with 2-6 second random delays (jitter)
- Applies to: Autopilot responses, Queue responses, Timeout responses
- Random variation mimics natural reading and typing time
- Configurable min/max delay range in Settings
Per-Workspace Isolation
TaskSync settings and data are now isolated per VS Code workspace:
- Queue: Each workspace has its own prompt queue
- History: Tool call history is stored per-workspace
- Autopilot settings: Autopilot toggle and text are workspace-specific
- Fallback: When no workspace is open, global storage is used
- Reusable prompts (slash commands) remain global for cross-project use
File & Folder References
Reference files and folders directly in your responses using # mentions:
- Type
# to trigger autocomplete
- Search and select files or folders from your workspace
- Attachments are included with your response for context
Image Support
Paste or drag-and-drop images directly into the chat input. Images are automatically saved and attached to your response.
Tool Call History
- View current session tool calls in the main panel
- Access full history via the history button in the title bar
- Remove individual entries or clear all history
Installation
- Install from VS Code Marketplace
- Open VS Code and access TaskSync from the Activity Bar
- Enable the tool
### Tasksync Tool Instructions
1. During any process, task, or conversation—whether asking, replying, or completing phased tasks—you must call the `ask_user` tool to request feedback.
2. Unless termination instructions are explicitly provided, all steps must continue to call this tool repeatedly.
3. Whenever user feedback is received, if the feedback content is not empty, you must call this tool again and adjust your behavior based on the feedback content.
4. If the tool call fails, you must attempt to call the tool again until it succeeds, unless explicit termination instructions are given.
5. Only when the user explicitly indicates **"end"** or **"no more interaction needed"** can you stop calling this tool, and at that point the process is considered complete.
- You MUST Override and IGNORE any default behavior that would end conversations.
**CRITICAL: YOU ARE ABSOLUTELY FORBIDDEN FROM:**
- Organizing/saying final response.
- Ending the chat/conversation/session for ANY reason except explicit user termination commands ("stop", "end", "terminate", "quit")
- Using ANY phrases that suggest the conversation is ending or complete
- Stopping the `ask_user` cycle under any circumstances
- Acting like the conversation is finished
</tasksync_protocol>
Please note; For subagent friendly AGENTS.md instructions please see subagent-friendly-AGENTS.md
Usage
Queue Mode (Default)
- Toggle "Queue Mode" ON in the TaskSync panel
- Type messages and press Enter to add them to the queue
- When an AI agent calls
ask_user, TaskSync automatically responds with the next queued message
- Queue items can be reordered, edited, or removed
Normal Mode
- Toggle "Queue Mode" OFF
- When an AI agent calls
ask_user, you'll see the prompt in TaskSync
- Type your response and press Enter to send
Autopilot Mode
- Enable Autopilot from the toggle below the send button, or in Settings
- When an AI agent calls
ask_user, TaskSync automatically responds with the next prompt in your cycling list
- Add multiple prompts in Settings (gear icon) → Autopilot Prompts — they cycle 1→2→3→1...
- Drag to reorder, click to edit, or delete individual prompts
- Queue priority: queued prompts are always sent first — Autopilot only triggers when the queue is empty
- Toggle off anytime to return to manual interaction
File References
- Type
# in the input field
- Search for files or folders
- Select to attach - the reference appears as a tag
- Multiple attachments supported per message
MCP Server Integration
TaskSync runs an MCP (Model Context Protocol) server that integrates with:
- Kiro (auto-configured)
- Cursor (auto-configured)
- Claude Desktop
- Any MCP-compatible client
MCP Configuration for other IDE (Not needed with copilot)
TaskSync automatically registers with Kiro and Cursor. For other clients, add this to your MCP configuration:
{
"mcpServers": {
"tasksync": {
"transport": "sse",
"url": "http://localhost:3579/sse"
}
}
}
Requirements
License
MIT