vscode-agent-todos

This MCP server exposes a memory-driven task management integration for a VS Code–centric AI workflow. It enables persistent task tracking across sessions, integrates with Copilot’s assistant features, and provides a structured task tree with statuses like pending, in_progress, and completed. The server pairs with tooling that presents a visual Todo tree, and it can auto-sync todos into Copilot instructions to keep the AI context aligned with your project plan. You can use the provided commands to fetch, update, or advance tasks, and leverage the VR of task decomposition to manage complex, multi-step workflows across coding sessions.

Prerequisites:

• Node.js (14.x or newer) and npm
• A compatible MCP client environment or tooling to connect to the MCP server

Install and run locally:

1. Clone the repository
2. Install dependencies npm install
3. Start the server node server.js
4. Connect your MCP client (the MCP runtime you use for your environment) to the server named 'vscode-agent-todos' using the mcp_config format described below.

Alternative (if distributing as a package):

• Ensure package.json starts the server entrypoint (e.g., node server.js) and run npm install && npm start.

Notes:

• If your environment uses a different entrypoint, adjust the args accordingly (e.g., ["dist/server.js"]).

Tips and common considerations:

• Environment variables: adjust DEBUG and LOG_LEVEL to control logging in your environment.
• If you encounter connection issues, verify that the server is reachable from your MCP client and that the port is open if applicable.
• The server is designed to work alongside VS Code’s Copilot and agent modes; ensure Copilot integration is enabled in your VS Code setup if you plan to leverage the automatic todos injection.
• You can customize Todo persistence and workflow state behavior by configuring the underlying memory model and task schema in your deployment.