CodexMulti is an independent open-source project and is not affiliated with or endorsed by OpenAI.
Only the latest released version receives security updates.
| Version | Supported |
|---|---|
| Latest release | Yes |
| Older releases | No |
Use GitHub Private Vulnerability Reporting. Open this repository's Security tab, choose Report a vulnerability, and submit the private report there. Do not open a public issue for a suspected vulnerability.
Include the affected version, impact, reproduction steps, and any safe supporting material. Do not include credentials, tokens, or complete copies of auth.json or config.toml.
We aim to acknowledge a report within 72 hours. We will coordinate validation and remediation privately, then announce the fix through a release.
- Credentials stay on the Mac.
- The bundled proxy listens on loopback only.
- CodexMulti has no telemetry or analytics.
Examples that are in scope include:
- Exposure of credentials or account data caused by CodexMulti.
- Proxy traffic becoming reachable beyond loopback.
- Authentication or routing behavior that crosses CodexMulti account boundaries unexpectedly.
- Unsafe changes to files or services managed by CodexMulti.
Examples that are out of scope include:
- The behavior of Codex itself when a user explicitly configures it with
danger-full-access. - Vulnerabilities in OpenAI or ChatGPT services that are not caused by CodexMulti.
- Unsupported modifications, repackaged builds, or operating systems other than Apple Silicon macOS 26 or later.
- Reports that describe only expected local access by the signed-in macOS user.