Company
About Vigil
Vigil is an independent runtime control plane for AI agents. We check prompts, retrieved documents, outputs and tool calls against your policies before they execute — without tying you to one model vendor or agent framework.
Who we are
- Product
- Vigil — AI TRiSM Runtime Guardrail & Observability Platform
- Operated by
- YG
- Operations
- Tel Aviv, Israel · cloud-hosted worldwide
- Payments
- Paddle.com acts as our Merchant of Record for paid plans
- Support & technical inquiries
- support@vigiltrism.ai
- Security reports
- security@vigiltrism.ai
Our engineering philosophy
Vigil was founded by security-focused software architects and independent entrepreneurs to close the gap between rapid AI adoption and production-grade runtime safety.
- Radical transparency. We publish our benchmark methods, detection rates and known limits. No vendor catches 100% of attacks.
- Defense in depth. Assume the model will sometimes be manipulated; enforce pre-flight checks and tool-argument restrictions where actions execute.
- Fail closed by default. When a check cannot complete, safety takes precedence over availability.
- Never train models on customer prompts, outputs or traces.
Your data
Pre-flight scans are evaluated in memory and allowed inputs are not stored. Execution traces you send are kept for your plan's retention window, with credentials scrubbed first. Details are in the FAQ and our privacy policy. To report a vulnerability, see our security page.