OpenAnalyst

Two years ago I watched a friend who runs a $2M DTC brand spend an entire Tuesday afternoon pulling reports.

She logged into Meta Ads. Exported to CSV. Logged into Shopify. Exported to CSV. Opened Klaviyo. Exported to CSV. Then spent two hours in Excel stitching it all together to answer one question: "Are we wasting money on this campaign?"

The answer was yes. She'd been over-spending for 11 days. Nobody caught it because nobody had the time to look.

That was the moment.

What we built

OpenAnalyst is a free desktop app. You download it, connect your tools, and tell it what you want in plain English. 42 AI agents handle the rest.

Not a chatbot. Not a thin wrapper around an LLM. Actual agents wired into Meta Ads, Google Ads, Shopify, Klaviyo, HubSpot, and 22 other tools through the MCP protocol. They read your data, draft the work, route it through a human reviewer, and ship.

The desktop app is free. No credit card. Your work stays on your machine. We also do custom enterprise builds — fixed price, two to four weeks, client owns the code. SOC 2 Type II. GDPR compliant. NVIDIA Inception member.

Numbers, because Indie Hackers deserves numbers

• 10,000+ campaigns powered since launch

• 14 custom builds shipped for clients

• 161 stars on our open-source VS Code extension (21 public repos on GitHub)

• 13 people on the team (India + SF)

• Free desktop app. Fixed-price enterprise builds.

• One DTC beauty brand: +34% recovered cart revenue in 90 days, 9 days from kickoff to production

• One B2B SaaS: lead-to-first-touch from 3 days → under 5 minutes

What we got right

Desktop over web. Most AI tools are browser tabs you have to log into. We built a desktop app that sits on your machine and works with your local tools. No data leaves your environment unless you connect a cloud service. Turns out privacy-conscious teams really care about that.

Free first. The desktop app is completely free. No credit card, no 14-day trial theater. People try it, it works, and enterprise teams come to us when they need custom builds. The free tier is our best salesperson.

Real integrations, not API wrappers. We didn't build a thin UI on top of OpenAI and call it a product. Each agent connects to real APIs — Meta, Google, Shopify, CRMs — and does actual work. It's harder to build, but it's also harder to copy.

Open source from day one. The VS Code extension and core agent framework are public. It builds trust with technical buyers and attracts contributors who later become customers. 161 stars isn't huge, but every one is a developer who understands what we're building.

What we got wrong

Too many things at once. We launched with 42 agents. Should have launched with 5. The breadth impressed some people but confused more. We're now packaging specific workflows (cart recovery, paid media, CRM) instead of overwhelming people with choice.

The website took too long to simplify. Our first landing page was dense — platforms, features, integrations, all at once. Took us months to strip it down to "one ask, the whole loop." Simple converts better. We knew that. We still overcomplicated it.

The one thing that surprised us

The enterprise deals didn't come from cold outreach. Every single one came from someone who downloaded the free app, used it, and emailed us asking "can you build this for our whole team?"

We spent zero dollars on sales in our first year. The product sold itself. The lesson: if your free tier actually works, your customers will ask you to charge them.

What's next

We're focusing on three things:

Making the desktop app stupidly easy. One-click setup for the top 5 workflows. No configuration. Just open and go.

Growing the open-source community. More docs, more examples, more contributors. The 161 stars should be 1,000 by year end.

Publishing more case studies. The cart recovery numbers get people's attention. We need 10 more like it.

If you're building marketing tools or running a team that's drowning in dashboards, I'd love to know: what's the one workflow that eats most of your week? Drop it below. I'll tell you honestly if an agent can handle it today or if it's still firmly in "needs a human" territory.