This simple prompt instructions helps you work more efficiently, reduce premium request usage, and allow you to give the agent new instructions or feedback after completing a task. This is similar to Interactive/Enhanced Feedback MCP.
TaskSync is a terminal-based task agent with direct terminal communication. Your coding agent will actively requests tasks or feedback through terminal command $task = Read-Host "Enter your task" or read -p "Enter your task: " task for macOS/Linux, executes tasks autonomously, and operates forever until you stop it.
demo.mp4
- Human-in-the-loop workflow - Provide feedback or new task, saving you premium requests on your AI IDEs
- Terminal-based agent interaction - Your AI becomes a persistent terminal agent that actively requests tasks
- Autonomous operation - Agent runs continuously requesting and executing tasks
- Never terminates automatically - maintains persistent operation until you explicitly say "stop", "end", "terminate", or "quit"
- Get the TaskSync Prompt: Copy or download the prompt from here.
Optional: Specs Workflow (Alternative structured approach): For users who prefer Kiro-style spec-driven development, you can use the Specs-Tasksync instead. This approach transforms ideas into structured requirements, design documents, and task lists before implementation. Simply provide the TaskSync or Specs Workflow file as context, then specify in chat which file your agent will follow.
- Initialize Agent: Provide the TaskSync v4 protocol file (
tasksync.md) or (specs-tasksync.md) as context to your AI IDE or agent and type in chat:Strictly follow TaskSync Protocol #tasksync.md or specs-tasksync.mdto activate. For copilot: use it as custom chat mode. - Agent Activation: The agent immediately becomes a terminal-based autonomous agent and announces initialization.
- Task Input: Agent executes
$task = Read-Host "Enter your task"orread -p "Enter your task: " taskfor macOS/Linux and waits for your input.
Note: Task must be entered as a single line. When pasting in terminal, click "Paste as One Line".
For GPT Models Use Tasksync MCP.
Because agent mode depends heavily on tool calling, it's recommended that you turn on "Auto Approve" in the settings. Note that this will allow the agent to execute commands in your terminal without asking for permission. I also recommend bumping "Max Requests" to 999 to keep the agent working on long running tasks without asking you if you want it to continue.
You can do that through the settings UI or via your user settings json file:
"chat.agent.maxRequests": 999It's best to keep the TaskSync session for 1-2 hours maximum since the longer the conversation, the more hallucinations may occur. Start it in a new chat session when needed for optimal performance. This works best as custom chatmode.
For users who prefer MCP (Model Context Protocol) Server integration for feedback-oriented development workflows:
This MCP server enables continuous feedback loops during AI-assisted development by letting users provide real-time feedback through a feedback.md file. It reduces costly requests and makes development more efficient by guiding the AI with feedback instead of speculative operations.
Key Features:
- 🔄 Continuous Review Feedback: Real-time feedback collection during development
- 🖼️ Media Processing: View images in the workspace
- 📝 File-based Communication: Uses
feedback.mdfor seamless interaction - ⚡ Efficient Workflow: Reduces premium requests through interactive feedback
Source code and setup instructions available at: TaskSync MCP Server
This version is using Version 3 that uses tasks.md for task input and log.md for updates. For VS Code users who prefer an integrated sidebar extension experience:
If you have issues installing the extension on other IDEs follow this guide: VSIX Installation Guide
Source code available in the TaskSyncExtension folder - this is an open source project.
Features:
- 🎯 Integrated Chat Interface: Send tasks directly through VS Code sidebar
- 🤖 One-Click Copilot Activation: Click/send for initialization
- 📁 Automatic File Management: Creates and manages files
- 📎 File Attachment Support: Reference files for context
- 🏷️ Task Categorization: Organize tasks by type
- 💬 Real-time Chat: View conversation history instantly
📋 View Full Changelog - See detailed release notes and version history
If you prefer older versions of TaskSync, you can access them directly:
TaskSync V3 (Web UI version added)
- V3 Protocol
- File monitoring system with web UI
TaskSync V2 (Enhanced)
- V2 Protocol
- Terminal monitoring system with enhanced session management
TaskSync V1 (Original)
- V1 Protocol
- Simple autonomous agent protocol with basic monitoring
The TaskSync community can be found on GitHub Discussions where you can ask questions, voice ideas, and share your prompts with other people. Contributions to TaskSync are welcome and highly appreciated.
⭐ Drop a star if you find this useful!