Back to home

Manifesto

What we believe and why we built this.

Own Your Agents

Most AI products work like this: your data goes to their servers, their models process it, you rent access to the result. Cancel your subscription, lose everything. Switch providers, start over.

We think that's backwards.

AI agents are going to become essential tools-like your computer or your phone. You don't rent those. You own them. The data on them is yours. When you switch, your stuff comes with you.

Agents should work the same way.

Why We Built This

I run AI agents for my actual work. Every day. For months now.

The agents themselves got good fast. But everything around them-monitoring what they're doing, letting them reach me when I'm not at my desk, coordinating multiple agents-stayed painful. I was cobbling together scripts, checking terminal windows, losing track of what was running.

I wanted a dashboard. Not a hosted AI that pretends to be my agent-my actual agents, running on my machine, showing up in one place where I could watch them work.

That's Beanstalk.

How It Works

Your agent runs on your machine. Full stop. We don't host it, we don't see your conversations, we don't store your files. The compute happens locally. The data stays local.

What we provide is the connective tissue: a gateway daemon that runs alongside your agent, a relay that lets your agent reach you on mobile or ping you in Slack, and a dashboard where everything shows up.

Messages pass through our relay in transit, but we don't store them. No logs, no databases, no training data. Once delivered, they're gone.

This architecture is deliberate. We couldn't monetize your data if we wanted to-we don't keep it.

No Lock-In

Beanstalk works with whatever agent you're running. Claude Code, Cursor, your own custom setup-doesn't matter. If it can emit events, it can connect to Beanstalk.

We're not trying to be the platform that owns your AI relationship. We're trying to be the tool that makes your existing setup work better.

If you leave, your agents keep running. You just won't see them in our dashboard anymore. No hostage data, no migration pain.

Indie Means Something

I'm not backed by VCs who need a 10x return. I don't have a board pushing me to maximize engagement or find ways to monetize user data. I'm one person building a tool I use every day.

That means I can make decisions that prioritize the product over growth metrics. I can keep things simple. I can say no to features that would compromise the architecture.

It also means this is smaller and scrappier than the alternatives. Fewer features, less polish, more focus. Trade-offs are honest.

Who This Is For

Builders. Tinkerers. People who run their own agents and want visibility into what those agents are doing. Developers who don't trust black-box platforms. Anyone who prefers to own their tools.

If you want a turnkey hosted AI assistant, there are plenty of options. This isn't that.

If you want infrastructure that stays out of your way while making your agents more capable, you're in the right place.


- Tommy