google-tasks-mcp
An MCP server for Google Tasks with full API support including task-list CRUD, task CRUD, reordering, and a diff_tasks change-harvester that detects hidden completions.
README
google-tasks-mcp
An MCP server for Google Tasks with the full API surface: task-list CRUD, task CRUD, move/reorder, and a diff_tasks change-harvester that tells you what happened since your agent last looked.
Built for agents that maintain a task mirror (project lists your assistant keeps in sync, completions you tick on your phone that the agent picks up later), but it works fine as a general Tasks connector.
Why another one
The existing options each miss something this use case needs:
- zcaceres/gtasks-mcp, the most-cited one, has no task-list operations at all: you can't create, rename, or delete lists, which rules out per-project lists entirely. The
gtasks-mcpnpm name is also a tombstone (the package was unpublished). - arpitbatra123/mcp-googletasks covers the full surface but isn't on npm, so it's clone-and-build, and auth means pasting an OAuth code back through a tool call.
- google_workspace_mcp does everything, plus all of Workspace, plus an OAuth 2.1 setup and a native-build gotcha on Windows. Overkill if you only want Tasks.
None of them deal with the API's nastiest quirk: tasks completed in the Google Tasks apps become hidden, so a naive query silently misses exactly the completions a sync agent needs to see. diff_tasks exists because of that quirk.
Install
# 1. One-time auth (see Google Cloud setup below first)
npx google-tasks-mcp auth
# 2. Register with your MCP client, e.g. Claude Code:
claude mcp add -s user gtasks -- npx google-tasks-mcp
Any MCP-capable client works; the server speaks stdio.
Google Cloud setup (one-time, ~15 minutes)
The Tasks API requires your own OAuth client. No verification, no billing.
- console.cloud.google.com → create a project.
- APIs & Services → Library → enable Google Tasks API.
- Google Auth Platform (consent screen): User type External (Internal is Workspace-only). App name + your email. Scope:
https://www.googleapis.com/auth/tasks(classified sensitive, not restricted: no security audit needed). - Credentials → Create credentials → OAuth client ID → Desktop app → download the JSON.
- Save it as
~/.config/google-tasks-mcp/client_secret.json(or pointGTASKS_MCP_CREDENTIALSat it). - The trap everyone hits: while the consent screen's publishing status is "Testing", refresh tokens expire every 7 days and you will re-auth weekly. Set publishing status to In production (skip verification; you'll click through a one-time "Google hasn't verified this app" interstitial: Advanced → Continue). Tokens then persist indefinitely.
- Run
npx google-tasks-mcp auth: a browser opens, you approve, the refresh token lands in~/.config/google-tasks-mcp/token.json. Done forever (revoking access or 6 months of disuse are the only expiries).
Environment variables
| Variable | Default | Purpose |
|---|---|---|
GTASKS_MCP_DIR |
~/.config/google-tasks-mcp |
Config directory |
GTASKS_MCP_CREDENTIALS |
<dir>/client_secret.json |
OAuth client file |
GTASKS_MCP_TOKEN |
<dir>/token.json |
Cached refresh token |
Tools
Task lists: list_tasklists, get_tasklist, create_tasklist, update_tasklist, delete_tasklist
Tasks: list_tasks (filters: completed/hidden/deleted, updatedMin, due bounds), get_task, create_task, update_task, complete_task, delete_task, move_task (reorder, re-parent, or move across lists)
Sync: diff_tasks(since, [tasklist]) returns everything that changed after an RFC3339 timestamp, grouped per list into completed / active / deleted. It sweeps every list unless you name one.
Design notes (API sharp edges, handled)
- Due dates are date-only. The API silently discards the time portion, and naive RFC3339 values can land a day off. Pass
YYYY-MM-DD; the server normalizes to UTC midnight. - App-completed tasks go hidden. Ticking a task in the Google Tasks app sets
hidden: true; a plain list call never sees it again.diff_tasksalways queries withshowCompleted + showHidden + showDeleted, so nothing is missed. - No sync tokens. Unlike the Calendar API, Tasks has no incremental sync token.
diff_tasksusesupdatedMin; keep a snapshot on your side and pass its timestamp. positionis read-only. Reordering only works throughmove_task(parent + previous). This includes subtask nesting.- There is deliberately no
clear_completedtool. The API'stasks.clearwipes completed tasks from a list, which permanently destroys the evidencediff_tasksdepends on. An LLM should not be able to call that casually. If you truly need it, the Tasks apps expose it in their UI.
Development
git clone https://github.com/justerlex/google-tasks-mcp
cd google-tasks-mcp
npm install
npm run build
node dist/index.js auth # one-time
node dist/index.js # stdio server
One TypeScript file, ~450 lines: src/index.ts.
License
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.