Move Fast, Not Break Things

It’s 3 hours away from launching AgentQL on Product Hunt. I’m sitting in our office in Palo Alto, watching our team prepare for the big release, and reflecting on the journey that brought us here.

Over the past 12 months, we’ve turned a wild idea—that AI agents could handle tasks for humans—into a production-ready framework. This framework now allows developers to create reliable web agents for automation, data scraping, and testing. But today, I want to share something that’s been just as critical to our success: the internal infrastructure we’ve built that enables our team to move fast without breaking things.

When we started Tiny Fish, we knew we were building an infrastructure that would be critical for our customers, who depend on our services to run 24/7. So, from day one, we invested in building a world-class dev infrastructure. It evolved alongside our product, and over the past two weeks, as we’ve sprinted toward our launch date and doubled the code throughput, its value has become even more apparent.

This infrastructure allows our team to commit code changes thousands of times per week, while still maintaining the quality and reliability that our customers expect. Without it, we wouldn’t be able to accelerate as reliably as we have. Here’s a look at some of the tools and practices that have become essential to our workflow:

1. Terraform

Managing our AWS resources efficiently has been transformative. With Terraform, every aspect of our AWS environment—from domain configuration to ECS fleet management—is handled through code. This eliminates the need for manual changes in the AWS console, reducing the risk of human error. Every infrastructure change undergoes mandatory code reviews and terraform plan verification, ensuring that once a pull request is approved and merged, the changes are automatically propagated to AWS. This process democratizes infra management across our team, removing bottlenecks and enabling faster iteration cycles.

2. Infrahouse

Adopting Terraform came with its challenges, particularly the initial integration cost, which can be a significant hurdle for teams with limited bandwidth. That’s where Infrahouse comes in—a powerful open-source framework developed and maintained by our DevOps lead, Oleksandr Kuzminskyi , which streamlines the Terraform integration process, drastically reducing the setup complexity. With Infrahouse, our team can manage AWS access and credentials more efficiently than ever before, eliminating the need for cumbersome AWS CLI commands. This has made infrastructure management not only more accessible but also more intuitive for everyone on the team.

3. Vercel Deployment Preview

Vercel's deployment preview is simply brilliant. For every pull request, it spins up a live instance of the website, allowing the team to review updated web pages in real-time. This significantly enhances the efficiency of our code review process, as designers, product managers, and engineers can all provide feedback directly within the previewed pages. The result? Countless UI and UX improvements are captured and implemented before they ever reach production. Integrating Vercel with GitHub Actions was a seamless experience, and it has become an indispensable part of our development workflow.

4. GitHub Actions

We've leveraged on Github Actions to automate a wide range of tasks—from verifying AgentQL client SDK against multiple Python versions, to spinning up Docker images for validation. Every pull request triggers these actions, allowing us to catch issues ranging from syntax errors to integration faults before they merge into the main branch. Over the past two weeks alone, we've executed 4,600 actions, saving hundreds of developer hours and ensuring higher product reliability across the board.

In addition to the tools that streamline our infrastructure, we also rely on a set of frameworks and tools that handle critical systems for us. These allow our team to stay laser-focused on our core product.

Sentry - Reporting and Alerts

We use Sentry for performance and error monitoring. It provides the data we need to dive into production issues. The integration was straightforward, and it offers flexibility to grow with our needs.

Clerk - User Management

We initially began building user management in-house but quickly realized that creating a secure and seamless user experience involved more complexity than anticipated. It was clear that continuing down this path would pull valuable engineering resources away from our core product. Clerk provided a comprehensive solution, covering everything from account life cycle to payment integration. This shift allowed us to focus on our primary goals while still ensuring a great user experience.

We've also been building tools around LLM operations, including benchmarking, prompt tracking, and response calibration. We're always open to mature solutions in this space as they emerge, so if you have any recommendations, please share.

If you're curious about our infrastructure journey and want to learn more, feel free to reach out. If there's enough interest, we’d be glad to host a meetup at our office to exchange ideas.

Now, back to getting everything ready for the launch. Would love your support on product hunt, we will be live at 12:01am PST.