This repository provides a set of tools and an SDK for connecting AI agents to Todoist, enabling them to manage tasks on a user's behalf. It targets developers building AI-powered applications or integrating conversational interfaces with task management, offering a seamless way for LLMs to interact with Todoist data and functionality. The primary benefit is empowering AI agents with the ability to create, find, and modify Todoist tasks, thereby automating workflows and enhancing productivity.

How It Works

The core approach involves exposing Todoist functionalities as callable tools, compatible with both direct integration into AI frameworks (like Vercel's AI SDK) and through a Model Context Protocol (MCP) server. The MCP server, runnable via npx, acts as a streamable HTTP service that adheres to the OpenAI MCP specification, facilitating broad compatibility. Tools are designed for reusability and to enable complete workflows rather than just atomic actions, striking a balance between LLM flexibility and efficient task management.

Quick Start & Requirements

Installation can be done via npm install @doist/todoist-ai for direct dependency use or npx @doist/todoist-ai to run the MCP server. Prerequisites include a TODOIST_API_KEY environment variable and Node.js. The project offers extensive documentation for setting up the MCP server (docs/mcp-server.md), tool design (docs/tool-design.md), local development (docs/dev-setup.md), and interactive widgets (docs/widgets.md). The primary MCP server URL is https://ai.todoist.net/mcp.

Highlighted Details

• Supports MCP Apps for rendering interactive UI widgets (e.g., task lists) directly within AI chat interfaces.
• Tools are designed to enable complete workflows, offering more than just basic CRUD operations for LLMs.
• Includes OpenAI MCP compatibility, with search and fetch tools optimized for OpenAI's requirements.
• Tools are reusable and can be plugged into various conversational AI interfaces beyond the MCP server.

Maintenance & Community

The project utilizes release-please for automated version management and package publishing, driven by Conventional Commits. The README does not detail specific community channels (like Discord/Slack) or notable contributors.

Licensing & Compatibility

The license type is not explicitly stated in the provided README content.

Limitations & Caveats

The project is described as being in its "early stages," with expectations for future tool additions and improvements. No specific limitations, bugs, or unsupported platforms are detailed.