Skip to main content
Glama

Things 3 + your AI.

A little help with everything on your list.

Find the task you forgot. Turn a rough idea into a project. Make room for what matters today.

Things 3 MCP connects your AI app to Things on your Mac. MCP is the connection that lets an AI app use other tools. You don't need to write code to use it.

Start with a question

Add more when you're ready

“What's on my list today?”

“Add a task to book the appointment.”

“Find my notes about the trip.”

“Plan a project for the move.”

“What deadlines are coming up?”

“Move this task into my weekend project.”

macOS · Things 3 · Read-only setup first · MIT licensed

Choose your setup

✦ Let an agent help

↗ Set it up yourself

Give a coding agent the prompt below. It checks your Mac, installs the tool, and configures your chosen app.

Follow four steps with copyable commands and settings.

Copy the setup prompt ↓

Open the manual guide →

You'll need: a Mac with Things 3 opened at least once, and an AI app that supports local MCP servers. Agent-assisted setup needs a coding agent with access to your Mac; a chat window alone may not be able to install software.

Related MCP server: Things MCP

Set it up with an agent

Copy the whole block into your coding agent. It will ask which AI app you want to connect.

Help me connect Things 3 to my AI app on this Mac.

Read this setup guide first:
https://github.com/Chevalicious/things3-mcp-gtd/blob/v2.0.1/SETUP-AGENT.md

Use the released things3-mcp-gtd==2.0.1 package.
If you cannot access the guide, explain what access is missing.

Ask which AI app I want to connect. Check for an existing installation
and preserve its settings. Install the candidate only if needed.
Configure local stdio with read-only access and verify a harmless read.
Don't start a duplicate service, change my tasks, or expose a network port.
Keep credentials out of chat and source control. Explain permission prompts
and tell me what I need to click or restart.

Read what the agent will do →

Your first conversation

Once connected, try:

“Use Things to show me what's on my list today. Don't change anything.”

A returned task list—even an empty one—confirms Things can be read. Your AI app may ask you to approve the tool call.

Can't connect? Find the matching symptom →

Ready to let it make changes?

Reading comes first. Creating and updating tasks are optional next steps:

  1. Enable Things URLs in Things → Settings → General.

  2. Remove the read-only setting and reload your connection to enable creating tasks and projects.

  3. To edit existing tasks, add your Things URL token directly to private local settings.

Follow the write-access guide →

Enabling writes lets your connected agent change Things when it calls a write tool. Review your AI app's tool approval settings before doing so.

Where your information goes

The server runs on your Mac, has no telemetry, and writes no log file. Information returned by tools goes to your connected AI app; a remote AI provider can receive it. Read-only access still shares task titles and destinations. Notes, tags, and checklist details are available when requested.

Reads use Things' private database format in read-only mode. Routine writes use the official Things URL scheme. Unofficial; not affiliated with Cultured Code.

Privacy · Terms

Review the upgrade notes. Existing managed HTTP services need a migration review before replacing their environment. If you used the legacy OAuth endpoints, review the token rotation guidance.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    Related MCP Connectors

    Related MCP Servers