🚀 Claude AI + Zoom Automation: Building an AI Meeting Agent

TL;DR

I built a system that connects Claude AI with Zoom to automatically:

• Summarize meetings

• Extract action items

• Trigger follow-ups

---

🧠 The Idea

Meetings create a lot of friction:

• Notes get lost

• Action items are forgotten

• Follow-ups take time

So I built a system where AI handles the entire workflow.

---

⚙️ Architecture

Zoom → Webhook → Backend → Claude → Action Engine → Tools

---

🔌 Zoom Webhook Example

import express from "express";

const app = express();
app.use(express.json());

app.post("/zoom/webhook", async (req, res) => {
  if (req.body.event === "recording.completed") {
    const url = req.body.payload.object.recording_files[0].download_url;
    await processMeeting(url);
  }
  res.sendStatus(200);
});

app.listen(3000);

---

🧠 Claude Processing

async function analyzeMeeting(transcript) {
  const res = await fetch("https://api.anthropic.com/v1/messages", {
    method: "POST",
    headers: {
      "x-api-key": process.env.CLAUDE_API_KEY,
      "Content-Type": "application/json"
    },
    body: JSON.stringify({
      model: "claude-3-opus",
      messages: [{
        role: "user",
        content: `Summarize and extract action items:\n${transcript}`
      }]
    })
  });

  return res.json();
}

---

📌 Structured Output

{
  "summary": "...",
  "decisions": ["..."],
  "actions": [
    { "task": "...", "owner": "..." }
  ]
}

---

📬 Follow-Up Automation

async function sendFollowUp(summary, actions) {
  console.log(summary, actions);
}

---

🤖 Why This Matters

This is more than automation.

It’s the shift from: AI as a tool → AI as an operator

---

🔥 Use Cases

• Devs → auto-create tickets

• Teams → auto-sync decisions

• Founders → scale across meetings

---

🧭 Future

AI agents that:

• Join meetings

• Make decisions

• Execute tasks

---

💬 Final Thought

The goal isn’t better meetings.

It’s meetings that do the work for you.