Post 1 of 5 — The Handoff series

There's a quiet consensus forming in software right now, and we think it's wrong.

The consensus goes like this: every product needs its own agent. Connect all your tools - Jira, Linear, Slack, your calendar, your docs - to our box, and our agent will do everything for you. Notion has one. Your CRM has one. Your help desk has one. Your notetaker has one. Each of them wants to be the place where the work happens.

We tried to build that too. And the closer we got, the more obvious the problem became.

The thing nobody says out loud about product-agents

To use one of these agents well, you have to hand it the keys to your whole stack. That means connecting tools that often sit behind different security reviews, different IT approvals, different access tiers. So the experience at the start is frequently: it doesn't quite work yet, or I'm limited because of what I'm allowed to connect. You bought a Ferrari and you're driving it around the parking lot.

And there's a deeper issue. You probably already have an agent you like. Maybe it's Claude Cowork. Maybe it's Codex. It runs locally. It's configured exactly how you want it - your VPN, your connections, your instructions for how you write, your service-type definitions, the whole setup you've quietly perfected over months. That's your agent. The switching cost to somebody else's is enormous, and most people simply won't pay it.

Someone put this perfectly in a post that's been making the rounds: I don't want to use your agent. And honestly? Fair. Every product now ships an agent. You don't know exactly what it'll do. Adopting it has to be worth it. Usually it isn't.

So we stopped trying to be your agent

Here's the reframe that changed how we think about Earmark: what if Earmark works with the agent you already have?

Bring your own agent. If you're going to use Claude Cowork, or Codex, or whatever harness you've already set up - Earmark becomes the tool of choice that feeds it. We integrate so deeply with what you've already built that there's no reason to look elsewhere for your meetings.

This isn't a compromise. It's better in almost every dimension that matters:

• No setup tax. You already configured your agent. We don't ask you to rebuild any of it inside our product.

• No IT bottleneck. We don't need to connect to your tools. Your agent already has the access it needs.

• No new cost center. The execution runs on the agent you (or your company) already pay for. We're not in the middle of that bill.

• No lock-in. You use two different local agents on any given day and four different meeting platforms in any given week. Everything in software is trying to lock you into one proprietary thing. We're not. We meet you where you already are.
  
  

Where Earmark fits

Think of it as a clean division of labor.

Earmark holds the context. Every meeting you've ever had - and, with shared projects, every meeting your team has had - captured, structured, and instantly queryable. That's the part nobody else does well, and it's the part that makes everything downstream good.

Your meeting becomes the prompt. We take that context and do the first pass of the work: draft the email, build the deck, write the tickets, generate the recap. Then - and this is the whole trick -we hand it off to your agent to actually push it into your tools. One button. It just runs your flows.

You keep the loop. You decide where the work happens. Locally, on the agent you trust? Or, if you'd rather, on a managed agent we run for you. Choose your own adventure. For the foreseeable future, most people are going to choose local - and that's exactly what we're built for.

Why this is the right bet

The labs are going to keep getting faster. Anthropic and OpenAI are shipping at a pace no notetaker can match. The instinct to compete with that - to be the everything-app - is a treadmill that gets more expensive every quarter.

We'd rather ride the wave than fight it. The better Cowork and Codex get, the better Earmark gets, because we're the layer that arms them with the one thing they can't generate on their own: what was actually said in the room.

Your meeting is the prompt. Your agent does the work. We make the two talk to each other.

Over the next few posts we'll show you exactly how - perfect memory, shared projects, the handoff itself, and the loops that mean you don't even have to ask.

Want early access? waitlist@tryearmark.com.