Task Manager MCP Server

The Task Manager MCP Server bridges the gap between advanced AI assistants and real‑world software development workflows. It empowers tools such as Cursor to harness the full potential of Gemini 2.5’s gigantic 1‑million‑token context window while keeping costs and usage within a modest subscription budget. By acting as an intermediary, the server extends an assistant’s “agentic” capabilities—allowing it to plan, execute, and refine complex feature development without being constrained by the limited context that Cursor normally imposes.

At its core, the server accepts a high‑level feature description and delegates to an LLM that performs recursive task decomposition. The resulting plan is stored as a JSON file in , ensuring that every subsequent interaction has full historical context. When the assistant needs clarification, a WebSocket‑based UI pauses planning and presents the user with an interactive task list. This clarification workflow keeps the conversation fluid, allowing developers to adjust the plan on the fly or review partial progress before proceeding. Once a feature’s original tasks are finished, an optional automatic code‑review step analyzes the latest and generates additional tasks if necessary, keeping quality assurance tightly coupled to development.

The Svelte UI is more than a dashboard; it is an integral part of the workflow. Developers can open the UI after each planning cycle, manually tweak tasks, and monitor real‑time progress. CRUD operations are exposed through simple API endpoints, making it trivial to add or remove tasks without leaving the editor. The server also exposes a “clarification” endpoint that can be called from Cursor or other MCP‑compatible clients, enabling seamless integration into existing toolchains.

For developers who rely on AI assistants to accelerate feature delivery, this MCP server offers several unique advantages. It eliminates the need for expensive proprietary models by leveraging OpenRouter‑compatible endpoints, and it guarantees an effectively unlimited context window regardless of the underlying editor’s limitations. The combination of recursive planning, persistent conversation history, and an interactive UI creates a robust, end‑to‑end solution that scales from small tweaks to large architectural changes. Whether you’re prototyping a new micro‑service or refactoring an entire codebase, the Task Manager MCP Server turns your AI assistant into a full‑featured project manager that keeps developers focused on code rather than context constraints.