Episode 16 of our new Series 'Code with AI the Right Way' is here! — and this time, we're building a Airtable Clone LIVE from scratch! (It even includes AI Agents and B2B Billing!) This is a LIVE build — mistakes, debugging, and all. That's the point. You learn more watching someone solve real problems in real-time than from a polished, pre-recorded tutorial.

Vercel Eve might be the new standard for building production-ready AI agents. Before this, building a real agent meant hand-rolling a mountain of plumbing: streaming, model routing, durable long-running workflows, safe sandbox execution, Slack/web channels, MCP connections, human-in-the-loop approvals, observability, scaling, and more. In this video, I break down Vercel’s full Agent Stack and show how Eve brings it all together with a filesystem-first framework for agents. We build and inspect a data analyst agent called Pulse, run it from the web, terminal, and Slack, look at sub-agents, schedules, Workflows, AI Gateway, Sandbox, Connect, and how it all deploys alongside a Next.js app. 🎯 What You'll Learn: ✅ Vercel Eve for building filesystem-first production AI agents! ✅ Vercel AI SDK + AI Gateway for streaming, model calls & model routing ✅ Vercel Workflows for durable long-running agent execution ✅ Vercel Sandbox for safely running agent code in isolated environments ✅ Skills, tools, sub-agents & schedules for more powerful agent behaviour ✅ Slack, terminal & web channels for talking to the same agent from anywhere ✅ Next.js integration with withEve so your app and agent deploy together + SO MUCH MORE!

Sending SMS, WhatsApp, and RCS messages should be simple — but in the real world, you end up dealing with registrations, approvals, carrier routing, country rules, retries, and webhooks just to answer one question: did the message land? This video shows how to avoid all of that complexity with Sent, a single API that can route messages across WhatsApp, SMS, and RCS automatically. Sent handles channel availability, formatting, fallbacks, compliance, delivery states, and webhook updates behind the scenes, so your app can send the right message through the best channel without maintaining separate integrations for each one. I’ll walk you through a full delivery tracking demo using the Sent's TypeScript SDK, Convex, real webhooks, approved templates, contacts, automatic routing, and AI agent setup with the Sent's MCP Server. We’ll cover: ✅ Why sending SMS gets complicated fast ✅ How Sent routes WhatsApp, SMS, and RCS from one API ✅ Setting up API keys, webhook secrets, and template IDs ✅ Sending real WhatsApp and SMS messages from a demo app ✅ Tracking queued, sent, delivered, and read states in real time ✅ Using templates, dynamic variables, buttons, and rich WhatsApp messages ✅ Handling Sent's webhooks with Convex or Next.js route handlers ✅ Using the Sent's MCP and docs to build faster with AI agents If you’re building apps that need reliable user messaging, this is one of the cleanest ways I’ve seen to ship multi-channel delivery without fighting every carrier and platform yourself.

Everyone wants their own team of AI agents now — one checking your tasks, one answering in Slack, one watching your portfolio, one building dashboards, one shipping code. But setting that up yourself with open-source agents can get scary fast. You need to understand servers, security, integrations, permissions, agent memory, triggers, and what happens when something goes wrong. In this video, I show you how Hyperagent lets you build that same kind of AI team without needing a Mac mini, VPS setup, 3am debugging session, or deep technical knowledge. We build multiple agents live, including a personal assistant that checks Notion, Gmail, and Linear, plus an investment analyst that can research updates, respond through email and Slack, use skills, create dashboards, and improve itself over time. We'll cover: ✅ Why local AI agent setups can be risky if you don't know what you're doing ✅ What Hyperagent is and why it feels like the simple version of running an AI team ✅ How to create your first agent using chat ✅ Building a personal assistant agent for tasks, emails, calendar, Notion, and Linear ✅ Building an investment analyst agent with skills, Slack, email, and dashboards ✅ How Hyperagent handles integrations with granular permissions ✅ Using live mode, email, Slack, Telegram, webhooks, and schedules to trigger agents ✅ Shared memories, knowledge, teams, projects, and agent self-improvement ✅ How developers can connect GitHub and let agents collaborate on code This is how you go from "I wish I had an AI team" to actually running one — without needing to become a security expert first.

AI coding agents are getting really good at writing code, but there’s still a big gap: a lot of real-world app setup happens outside the codebase. Things like authentication, environment variables, organizations, billing, plans, production config, permissions, and dashboard setup usually still require a human to jump between tools, copy keys, configure services, and connect everything manually. That’s where CLI tools become a game changer. In this video, I show how giving an AI coding agent access to the right CLI, skills, and project context lets it do way more than just generate code. Instead of only relying on MCP tools or local project files, the agent can start interacting with real services from the terminal and handle setup steps that previously fell back to us. For this demo, we use the Clerk CLI as the example. We start with a blank Next.js app, authenticate the CLI, run the setup flow, install Clerk’s agent skills, and then push the workflow further by asking an AI agent to build a full multi-tenant SaaS app with authentication, organizations, billing plans, feature gating, and an upgrade flow. The bigger idea is simple: as more platforms expose powerful CLIs, AI agents become much more useful. They can write code, configure services, set up products, connect infrastructure, verify the result, and help us ship full applications faster. We’ll cover: ✅ Why CLI tools unlock a new level of power for AI coding agents ✅ Where MCP tools and skills help, and where CLIs take things further ✅ How agents can now handle setup work we previously had to do manually ✅ Starting from a blank Next.js app ✅ Using the CLI to configure authentication from the terminal ✅ Installing agent skills so the AI has better product context ✅ Giving Claude Code access to CLI-powered workflows ✅ Creating a working sign-in and sign-up flow ✅ Setting up organizations for a B2B SaaS app ✅ Creating billing plans, pricing, and feature gates ✅ Testing subscriptions, organizations, and upgrades