Qodo builds AI tools for code review and software quality with the aim of giving coding agents access to company knowledge needed to understand the code they change. Co‑founder and CEO Itamar Friedman describes the solution as “the coding agent’s best friend.”
Purpose of the plugin
Qodo’s Codex marketplace plugin injects organizational rules, codebase knowledge and an independent review step into developers’ Codex sessions. The feature is designed for environments that can span many repositories — Itamar gives the example of 10,000 or more — where a local change can break downstream services even when the change appears correct in isolation.
Qodo uses OpenAI models, including GPT‑6 Astra, to gather engineering knowledge, review code and monitor software quality. Before Codex starts writing code, the plugin can retrieve relevant codebase rules, pull request history and common patterns. After Codex makes changes, Qodo independently reviews those changes and returns findings into the same session, before the developer opens a pull request.
How it reduces missed critical bugs
With the plugin, Codex can account for dependencies across repositories, including those hosted on different Git providers or connected through Kafka. Itamar says, “Qodo will make that connection and will help Codex look at the software as a whole and not miss those P0 bugs.” The intent is to surface high‑severity (P0) issues earlier in the development cycle.
Agents, roles and workflow
Nnenna Ndukwe, Qodo’s AI Developer Relations Lead, reports that developers are using this workflow with GPT‑6 models in Codex to build features and address technical debt. They automate the handoff between Codex’s code generation and Qodo’s independent review.
Itamar characterizes coding and review agents as “yin and yang”: one produces code, the other challenges it. Each role requires its own configuration and context and may involve multiple agents. His vision is for agents to discuss findings and “surface only what they don’t accept” for human review, so developers act as planners and final reviewers who can examine unresolved issues alongside the agents’ discussion, context and prior developer decisions.
“That’s how you create trust,” Itamar says. "I think it makes the whole process overall more efficient," Nnenna adds. Qodo’s approach combines the strengths of AI and traditional engineering: company knowledge guides coding agents, independent review verifies their work, and humans resolve remaining questions.
Technical augmentations and decisioning
Internally, Nnenna has used GPT‑6 Astra’s Extra High and Ultra effort settings in automated development workflows and knowledge work. Itamar also highlights OpenAI’s recently launched Decisions API as a tool Qodo can use to determine which rules and agent skills should apply at each stage of development, rather than delegating every selection to a language model.
For example, when 200 rules could be relevant, the Decisions API could preselect which ones need to be checked, so the language model reasons only over the narrowed set—speeding the process, cutting cost and improving review quality. “The Decisions API is an immense opportunity for us to bring more trust to the process,” Itamar says.
A software‑factory vision
Itamar describes Qodo’s approach as the foundation for a software factory: a system customers could assemble and own that brings coding and review agents together with an understanding of how software changes in production. Working with OpenAI, Qodo aims to put more of a company’s institutional experience to work through coding agents while letting developers focus their judgment where it matters most.
The plugin and agent‑driven workflow are intended to reduce the risk of critical bugs reaching production, accelerate development cycles, and increase developers’ confidence in automated tools.
