SecretOps, Evolved

You have a new project and you want to build a web application. What do you do? At its simplest, you create a git repo, add a few dependencies like node and Next.js, push it up to GitHub, and deploy the application to a PaaS like Vercel. If that application needs to do stuff, like send emails, or receive payments, you’d sign up for services like Resend or Stripe, generate API keys, save them into a dotenv file, and paste those values into the PaaS where it gets deployed alongside your app. If someone else joins the project, you send them a copy on Slack. These API keys are sensitive credentials called Secrets, and this activity, at its most basic level, is called “SecretOps”¹.

SecretOps has fundamentally been the same for over a decade, but the burden of doing it has increased. Integrations, tools, and SaaS products have grown so numerous, your typical early product utilizes over a dozen of them, and managing and sharing their secrets has become quite unwieldy. A new term, “Secrets Sprawl”², was coined to describe this unwieldiness, and with the introduction of solutions like HashiCorp’s Vault, a new category, Secrets Management, emerged. The Secrets Management market today is quite stratified. Vault, which sits at the top, has evolved into a complex, infrastructure-heavy, enterprise solution. Newer, middle-market players like Infisical and Doppler emerged with simpler, more turn-key solutions to counterweight Vault’s complexity and cost.

But today, there is a shift in the world brought forward by AI. The builder in that story is no longer working alone. There is now an AI agent building along side, and it can write code, run commands, tweak infrastructure, and ship deployments all on its own. But the moment it needs a Stripe key, it has to stop and ask you. It doesn’t matter whether that key ends up in a dotenv or a Terraform variable, someone still had to go get it. Every other part of that first story the agent can now do by itself. We are fixing the latter. If that sounds exciting, we want to talk to you.