Security at kill9
Your code is your most valuable asset. We treat it that way.
Data encryption
All data encrypted in transit (TLS 1.3). Provider credentials are stored locally in a plaintext auth.json file written with mode 0600 (owner read/write only). All local files are protected with restrictive permissions. Secrets never leave your machine unencrypted.
No training on your code
Your source code is never used to train models. It is processed solely to execute agent tasks you request.
BYOK model access
Bring your own API keys. Your code goes directly to your chosen model provider — we don't proxy or store prompts.
Security-first architecture
IPC sandboxing, zone-based file access control, encrypted credential storage, shell injection prevention, and input validation across all system boundaries.
Role-based access
Fine-grained permissions at the organization, workspace, and repository level. Organization and workspace actions generate audit logs.
On-prem deployment
Enterprise customers can deploy kill9 on their own infrastructure. Code is processed locally — only model inference calls leave the device, routed to your configured provider.
Found a vulnerability? Report it to security@kill9.ai. We respond to all reports within 24 hours.