Greptile builds AI agents that review and test code changes with full context of the codebase. Co-founder and CEO Daksh Gupta says the company is in the business of reducing bugs and making the world’s software more reliable, which means testing an application’s behavior as well as reading its code. GPT-6 Astra’s computer use gives Greptile a way to automate that testing at scale.

For pull requests, Greptile can start a development server, resize the browser to an iPhone-sized viewport, and simulate a user moving through the application. In Daksh’s example, a button disappears when a website is viewed on an iPhone; the model’s visual understanding lets it inspect the screen and spot the hidden button. He argues that a QA firm testing all changes is impractically slow compared with how fast code is shipped with these models, so GPT-6 Astra is used to automate application testing at scale.

Greptile also addresses the security risk from more powerful models, which Daksh says make it easier for attackers to exploit applications and put even previously non-typical companies at risk. Daybreak is presented as a safe way to make it easy for even a startup to secure its systems, with customers having more sophisticated defense mechanisms than adversaries have offense mechanisms.

On the developer side, Greptile’s Knowledge Base indexes a company’s codebase and documents how it works in a form people and agents can read. It exposes that documentation to Codex through a Model Context Protocol (MCP) server, letting Codex write code that follows the customer’s existing patterns. Daksh says thousands of customers have connected Codex to Greptile’s knowledge base, and the team tracks how often AI-generated code is reverted: with context, fewer reverts happen, and customers save time and inference, putting those savings toward new work. Greptile itself is a heavy Codex user across every internal function, especially engineering.

Looking ahead, Daksh says the only real bottleneck is how quickly the team can turn ideas into working software. Model intelligence is already sufficient for many tasks, so Ultrafast is meant to speed up development. The longer-term goal is to make changing and shipping critical enterprise software as easy as shipping a side project, with OpenAI helping Greptile build the validation that requires.