How CodeOwl protects your codebase through ephemeral execution, zero-training commitments, and defense-in-depth isolation.
Zero Code Storage
Diffs processed in memory and wiped immediately
Zero Model Training
Your code is never used to train or fine-tune models
Least-Privilege GitHub App
No admin scopes, branch write, or secret access
CodeOwl was engineered from day one on a zero-persistent-code paradigm. We treat customer source code with the highest sensitivity:
Your code belongs exclusively to you and your team. Under no circumstances is customer code or PR discussions used to train, evaluate, or improve any public or foundational AI models.
CodeOwl operates as an official GitHub App rather than an OAuth user token or Personal Access Token (PAT). This provides critical security advantages:
| Permission | Level | Security Rationale |
|---|---|---|
| Pull requests | Read & Write | Required to read diffs and post review comments inline |
| Repository contents | Read-only | Used strictly to fetch import references and .owl/rules.yml |
| Checks | Read & Write | Reports pass/warning/fail checks back into the GitHub PR UI |
| Metadata | Read-only | Default GitHub permission for identifying repo name and owner |
We never request repository administration, billing, secret management, or commit push permissions to source branches.
Our review execution cluster runs in modern cloud environments with strict network segmentation, automated DDoS protection, and continuous vulnerability scanning.
We take security issues seriously. If you believe you have discovered a vulnerability in CodeOwl or its GitHub App integration, please disclose it responsibly:
Direct Security Contact:
Neelesh Kurmi
Email: neeleshp212@gmail.com
Initial Response SLA: Within 24 hours
Please include detailed reproduction steps, the affected endpoints, and sample payloads. We ask that you do not publicly disclose vulnerabilities before a patch has been released.
© 2026 CodeOwl · Built with care by Neelesh Kurmi