Creativity · MCP — server

MCP Todoist Server

The MCP Todoist Server uses the Todoist REST API to add, update, reschedule, and complete tasks. MCP clients can query projects and filters, apply labels, and set due dates with natural-language parsing. It's a favourite 'weekend project' MCP server for learners building their first personal-productivity agent.

MCP facts

Kind: server
Ecosystem: anthropic-mcp
Language: TypeScript / Python
Transports: stdio

Capabilities

Tools: add_task, complete_task, reschedule, list_tasks, list_projects, filter_tasks
Resources: todoist://project/{id}, todoist://task/{id}
Auth: Todoist personal API token

Install

npx -y @community/mcp-server-todoist

Configuration

{
  "mcpServers": {
    "todoist": {
      "command": "npx",
      "args": ["-y", "@community/mcp-server-todoist"],
      "env": {
        "TODOIST_TOKEN": "${TODOIST_TOKEN}"
      }
    }
  }
}

Frequently asked questions

Is it safe to give an LLM write access?

Todoist has no destructive operations the average user can't undo, but it's still wise to start with read tools and add write ones behind approval prompts.

Why is it a good learning project?

The API is compact, well-documented, and covers realistic scenarios (projects, due dates, labels). Wrapping it in MCP teaches the fundamentals of tool schemas, OAuth-free tokens, and idempotent operations.

Does it handle natural-language due dates?

Yes — Todoist's `due_string` field accepts phrases like 'tomorrow at 5pm', so the LLM can forward user language directly instead of computing timestamps.

Sources

Todoist REST API v2 — accessed 2026-04-20
Model Context Protocol — accessed 2026-04-20
MCP servers repo — accessed 2026-04-20