Key takeaways
- Scan a whole repository, chosen folders or just your recent Git changes for security vulnerabilities.
- It goes beyond finding: it can validate candidate findings, generate patches and verify existing fixes.
- Results can be exported as SARIF, JSON or CSV, or tracked in Linear, Jira or GitHub.
- It is open source under Apache-2.0, and local scans work without the optional cloud connection.
- The catches: it needs Node.js and Python, cost limits are estimates, and some findings need extra access.
What it is
Codex Security is OpenAI's command-line tool and TypeScript SDK for finding, validating and fixing security vulnerabilities. It is open source under the Apache-2.0 licence, and the code is on GitHub at openai/codex-security.
The problem it solves
Finding a vulnerability is only the start. Developers and small security teams then have to work out whether a report is real, how serious it is and what to change. The project aims to cover that whole path, from the first scan to a fix, in one workflow.
It can be used inside Codex, from a terminal, or automated through the CLI and TypeScript SDK. The suggested prompt asks for the saved report, source evidence and coverage limits, so the output shows what was and was not checked.
What you get
- Flexible scope: scan repositories, selected paths or Git changes, so a quick check of a branch does not need a full run.
- Fewer false alarms: it can validate candidate findings, generate patches and verify existing fixes.
- Less noise: deep scans repeat standard scans and return their deduplicated findings.
- Paperwork help: it can draft SECURITY.md policies and save threat models for later review.
- Fits your tools: findings can be tracked in Linear, Jira or GitHub, or exported as SARIF, JSON and CSV.
The catches
- Requirements: it needs specific Node.js versions and Python 3.10 or newer.
- Access: some cybersecurity requests and protected findings require Trusted Access for Cyber.
- Cost: the example workflow uses an estimated $5 limit per scan, and requests already in progress can exceed the limit.
- Maturity: the local serve command and browser dashboard have been removed, so check the documentation before relying on older guides.
- Providers: scans support OpenAI and several other model providers, which means your code is processed by whichever one you choose.
What we did not test
We did not run Codex Security, sign in, scan any code or check how accurate its findings or patches are. Every claim here comes from a literal quote in the project's README or the GitHub API, and the feature descriptions are the project's own.
Verdict
Editors' view: this is aimed at developers and small teams who want security checks in their everyday workflow, especially if they already use Codex or GitHub. If you handle sensitive code, read the provider and access terms first, and try it on a non-critical repository before trusting its output.
Sources
Need help adopting this, or have a software project in mind?
Tell us what you are building. The initial consultation is free. Email [email protected] or follow @zolnaltd.
Start a conversation →