@PrajwalTomar_: I built this in ONE hour. A discovery call ends, Granola catches the notes, Lovable turns them into a scoped project. →…

I built this in ONE hour.

A discovery call ends, Granola catches the notes, Lovable turns them into a scoped project.

→ Every requirement extracted with the exact quote from the call → Scope, timeline, and pricing drafted automatically → Follow-up email ready before the client is back at their desk

The agency that sends the scoped proposal first usually wins the client. Ours now goes out while the competition is still transcribing the call.

That’s the real unlock of Connectors. Lovable can now see the tools your business already runs on.

Full breakdown of my 2026 system below.

Bookmark this.

P.S. The screen recording uses mock call data so I can show the app without exposing client info.

I genuinely did not expect this to work this well.

I hired an AI employee at the agency this week. It runs on an open-source model.

I built a client research agent on Hyperagent and picked Kimi K2.6 from the model list. Now I drop a company link into Slack and it goes to work on its own: opens a browser, digs through their site, and comes back with a full pre-call briefing.

Then it publishes a beautiful briefing page for the prospect. Funding timeline, gaps we can pitch, 5 intake call questions, one red flag. Ready before the call even starts.

The wild part is the model picker. GLM 5.2, DeepSeek V4, Kimi K2.6, and Qwen 3.7 are all live on Hyperagent now, right next to the frontier models. Swap the brain under the same agent anytime and compare.

Watch it work through a real prospect in the video.

Open-source caught up faster than anyone expected.