• This New Way
  • Posts
  • AI agents run your inbox, calendars, and socials

AI agents run your inbox, calendars, and socials

AI agents that act are here. Learn how to build them.

AI is evolving from chat to action.

In this episode of This New Way we sat down with Sam Partee, co-founder of Arcade, the toolkit powering agents that do real work. From autonomous social media workflows to AI-managed inboxes, Sam walks us through how he’s building smarter, safer, and more hands-on agents.

You’ll want to forward this one to your Engineering teams.

Read the step-by-step tutorials and AI tools mentioned below.

Brought to you by Fellow - the most secure AI Meeting Assistant

Tutorial 1: Build an AI agent that handles your inbox

With the right AI tools, you can automate your inbox without hiring another assistant.

1. Set up an Inbox Agent
Arcade lets you create “agents” – AI bots that can log into Gmail, read your emails every hour, label what matters, and even draft replies.

2. Add an iMessage and Calendar Agent
Want it to handle even more? You can connect AI to your iMessage and calendar too. For example, you can use Claude (an AI model) to send texts, confirm meetings, and follow up with customers.

Why it works:
This isn’t some theoretical AI. These agents actually log in with your permissions, pull past context, and take action—on your behalf.

Tutorial 2: Use AI to speed up engineering

Busy teams need more than autocomplete – they need AI that understands their codebase and helps everyone move faster, without cutting corners.

Here’s how to get started with Cursor, the smart AI coding assistant.

1. Download and Install Cursor

  • Visit cursor.so and download the editor for your OS.

  • Cursor is built on VS Code, so it’ll feel familiar to most developers.

2. Connect Your Codebase

  • Open your project in Cursor.

  • Cursor will index your full codebase using vector embeddings, so it can understand structure and context.

3. Ask Cursor to Edit, Refactor, or Explain Code

  • Highlight code and type commands in natural language:

    • “Refactor this function to be more readable.”

    • “Add logging here.”

    • “Write a test for this module.”

  • Cursor uses its knowledge of your code to respond intelligently – no need to write every prompt from scratch.

4. Use Cursor as a Team-Wide Tool

  • PMs and QA can also use Cursor to:

    • Search and explore the codebase.

    • Understand features or debug flows.

    • Draft test cases or review AI-suggested edits – without pinging a dev every time.

5. Add Guardrails to Stay in Control

  • Use Cursor’s safety features like:

    • Cursor Rules – Set guidelines for how AI should behave (e.g., don’t touch auth logic).

    • Locked Files – Prevent AI from editing critical sections of code.

Pro Tip: Start small. Try Cursor on a non-critical repo first to see how it handles suggestions, then expand across your workflows.

AI Tools Mentioned

  • Arcade — Agent tool library with secure action + auth layer

  • Cursor — AI-native code editor with project context and agentic workflows

  • Claude — Front-end for natural language + tool calling

  • LangGraph — Agent orchestration and human-in-the-loop workflows

  • Twilio — For SMS reminders

  • TailScale — Secure network to access your laptop remotely

  • GitHub Copilot — Used in tandem with Cursor for inline coding

  • Vector databases — Enable semantic search across massive codebases

Why this matters:
If AI can automate busywork—email, docs, scheduling, and code cleanup—you can spend more time where it counts: building products and talking to users.

👉 Subscribe to This New Way: YouTube Channel

📩 Was this email forwarded to you? Subscribe here.

Until next time,

Aydin Mirzaee

CEO at Fellow AI and Host of “This New Way” podcast