Todoist MCP Server
Connect AI agents to Todoist tasks, projects, labels, and due dates via the official MCP server.
Updated
What is Todoist MCP?
The Todoist MCP server is a Model Context Protocol server that lets AI agents read and manage your Todoist tasks, projects, labels, sections, and due dates on your behalf.
It is officially built and hosted by Doist, the company behind Todoist. Rather than running anything locally, you connect to the remote endpoint at https://ai.todoist.net/mcp, a Streamable HTTP service. The first time an agent uses a Todoist tool, an OAuth flow runs in your browser so you grant access to your own account — there is no manual API-key setup for the hosted server.
Once connected, the server exposes tools for the full task-management workflow: adding and rescheduling tasks, finding tasks by date or filter, organizing projects and sections, managing labels and reminders, and pulling productivity and project-health insights. The same project is open source as @doist/todoist-ai on GitHub, so it can also be self-hosted over stdio with a Todoist API token if you prefer to run it yourself.
Tools the Todoist MCP server exposes
| Tool | What it does |
|---|---|
| add-tasks | Create one or more tasks, with content, due dates, priorities, and project/section placement. |
| update-tasks | Edit existing tasks — content, due date, priority, labels, or assignment. |
| find-tasks | Search and list tasks matching content, project, label, or filter criteria. |
| find-tasks-by-date | Retrieve tasks due on or around a given date. |
| complete-tasks | Mark one or more tasks as completed. |
| reschedule-tasks | Move tasks to new due dates. |
| add-projects | Create new projects to organize tasks. |
| add-labels | Create labels for categorizing and filtering tasks. |
| add-reminders | Set reminders on tasks. |
| get-productivity-stats | Return productivity statistics such as completed-task trends. |
Connect the Todoist MCP server
Claude Code
- 1
Add the remote server
Run: claude mcp add --transport http todoist https://ai.todoist.net/mcp
- 2
Authorize via OAuth
On the first Todoist tool call, a browser window opens — sign in and grant access to your Todoist account.
- 3
Verify
Run /mcp in Claude Code to confirm the todoist server is connected, then ask the agent to list your tasks.
Alternatively, install the official Todoist plugin: /plugin marketplace add doist/todoist-mcp then /plugin install todoist@doist.
Cursor
- 1
Open MCP settings
In Cursor, go to Settings > MCP and add a new server.
- 2
Configure the remote endpoint
Add an entry pointing to https://ai.todoist.net/mcp using HTTP transport. If your client only supports stdio, use: npx -y mcp-remote https://ai.todoist.net/mcp
- 3
Authorize via OAuth
Trigger any Todoist tool; complete the browser OAuth flow to link your account.
Sample use cases
An agent turns incoming emails or messages into Todoist tasks with due dates.
Action items are captured automatically without manual entry.
A morning agent reviews overdue tasks and reschedules them.
Your task list stays current and realistic each day.
An agent summarizes a project's open tasks and productivity stats.
Quick status visibility without opening the app.
Security & permissions
The hosted server at https://ai.todoist.net/mcp authenticates with OAuth: on first tool use, a browser flow asks you to authorize access to your Todoist account, so no API key is stored in your client config. The agent can then read and modify tasks, projects, labels, sections, reminders, and comments in the account you authorize — treat it as full read/write access to that workspace. If you self-host the open-source @doist/todoist-ai server instead, it authenticates with a TODOIST_API_KEY token you supply via environment variable; keep that token secret.
Using Todoist MCP with Gamut
In Gamut, you connect Todoist once and then reference it as a tool inside an automated agent workflow — Gamut manages the MCP connection and OAuth for you. A scheduled or event-triggered agent can, for example, create tasks from incoming requests, reschedule overdue items each morning, or summarize a project's open work and post it elsewhere, all by calling the Todoist MCP tools as part of a larger multi-step run. The agent acts within the scope you authorize on your own Todoist account.
Frequently asked questions
What is the Todoist MCP server?
It is Todoist's official Model Context Protocol server, maintained by Doist. It lets AI agents and assistants like Claude create, find, update, and complete tasks and manage projects, labels, and due dates in your Todoist account on your behalf.
Is the Todoist MCP server free?
Yes. The MCP server is free to use with a Todoist account and is open source (@doist/todoist-ai on GitHub). Todoist's standard free and paid plans apply to your account as usual.
How do I connect Todoist MCP to Claude or Cursor?
For Claude Code, run claude mcp add --transport http todoist https://ai.todoist.net/mcp and complete the browser OAuth prompt on first use. For Cursor, add https://ai.todoist.net/mcp as an HTTP MCP server (or use npx -y mcp-remote https://ai.todoist.net/mcp), then authorize via OAuth.
Is the Todoist MCP server hosted or self-hosted?
Both. Todoist officially hosts a remote server at https://ai.todoist.net/mcp that uses OAuth, so there's nothing to run locally. You can also self-host the open-source @doist/todoist-ai server over stdio using a Todoist API token if you prefer.