Product documentation
Security Model
Syndicate Code provides a verifiable boundary between untrusted AI proposals and authoritative code execution.
The Enforcement Boundary
The security boundary is defined at the point of action. The control plane intercepts every system call, file modification, and network request proposed by the AI runtime. If an action does not have a valid, time-bounded permit issued by the policy engine, the action is blocked.
Key Controls
- ✓
Capability Isolation
Policies define what an agent is permitted to do, independent of the model's internal instructions.
- ✓
Explicit Checkpoints
High-impact actions (e.g. repo push, production DB write) require out-of-band human approval.
- ✓
Tamper-Evident Audit
Cryptographically chained logs ensure that once an event is recorded, it cannot be altered without detection.
Contact & Disclosure
For security inquiries, vulnerability disclosures, or technical diligence requests, please contact our security team. We maintain a coordinated disclosure policy for all discovered vulnerabilities.