Ramp Deploys Codex with GPT-5.5 for Autonomous Code Review
Ramp's engineering team uses OpenAI's Codex to deliver pull request feedback in minutes, reducing manual review cycles and supporting internal agent development.
Last verified:
Code Review Feedback in Minutes, Not Hours
Ramp’s engineering team is deploying Codex with GPT-5.5 to replace multi-hour code review cycles with automated feedback delivered in minutes. According to the OpenAI Blog, the deployment has become a mandatory component of Ramp’s pull request workflow, with engineers specifically requesting Codex’s analysis by name. Austin Ray, who leads AI Developer Experience at Ramp, credits Codex with catching defects that escape both human reviewers and competing AI code-review systems.
The speed gain is material: engineers who previously waited hours for an initial review now receive substantive feedback immediately upon opening a pull request. Codex achieves this by reasoning deeply against the codebase rather than applying surface-level heuristics, according to Ray, delivering “a level of thoroughness that most human reviewers don’t have time for.”
Reasoning Depth as Competitive Differentiator
What distinguishes Codex from other automated review tools is its ability to navigate complex codebases without losing context. According to the OpenAI Blog, Ray notes that Codex’s reasoning capabilities reduce mental load in ways that would otherwise require “a ton of mental effort, a lot of sleep, and a lot of single-minded focus” from human engineers. This is particularly valuable in domains where business logic and interdependencies create high cognitive friction.
The tool accommodates different engineer preferences: command-line users can invoke Codex from the terminal, while those preferring visual interfaces use the Codex app, which provides guided navigation that improves engineering workflow velocity, according to Ray.
Internal Agent Development
Ramp’s AI DevEx team is also leveraging Codex to build internal tooling. According to the OpenAI Blog, Ray is using Codex to support development of an on-call assistant tool for Ramp engineers—though the available source material does not contain detailed specifications on that project’s scope or current status.
Why This Matters
For engineering organizations managing high-velocity development cycles, the shift from synchronous human code review to asynchronous AI-assisted review directly impacts time-to-merge and engineer throughput. Ramp’s public adoption signals confidence in Codex’s reasoning depth, which matters for enterprises evaluating whether to replace or augment human review gates. The move also demonstrates a use case where AI agents (Codex) are being used to accelerate development of other AI agents (On-Call Assistant), suggesting emerging patterns in how companies layer AI tooling within existing development stacks.
Frequently Asked Questions
How much faster is Codex code review compared to manual review at Ramp?
According to the OpenAI Blog, Ramp engineers now receive substantive pull request feedback in minutes instead of the hours they previously waited for a first manual review.
What makes Codex's code review different from other AI code reviewers?
According to Ramp's Austin Ray, Codex performs deeper reasoning against the codebase, delivering a level of thoroughness that typical human reviewers cannot match due to time constraints.
How does Codex support Ramp's internal tool development?
The source indicates Codex is being used to develop On-Call Assistant, an internal agentic tool, though details on that project are limited in the available excerpt.