Security & Safety

Syngentix is designed to be safe by default. Every AI request is screened before it runs, your data stays isolated to your workspace, and your connected credentials are stored encrypted. This page explains the protections that apply to your automations and agents.

In this guide

Content guardrails

Every message your agents send or receive passes through a guardrail check before it reaches an AI model. Both what goes in and what comes back are screened, so unsafe content is caught in either direction.

• Screens for sexual content, unethical or dangerous guidance, discriminatory content, and abuse.
• When something is blocked, the agent responds with a polite, professional refusal and offers a safe alternative.
• Guardrail screening can be toggled per agent to suit your use case.

Workspace & data isolation

• Your agents, workflows, files, and conversations stay within your workspace, governed by member roles.
• Access is scoped by role — Owner, Admin, Editor, and Viewer — so people only see and do what they should.

Encrypted credentials

• OAuth connections and secrets are stored in an encrypted credential vault, scoped to your workspace.
• Editors use connections to build and run; only Owners and Admins manage them.

Safe, predictable runs

• AI actions are checked against your available credits before they run, so spending stays predictable.
• Empty results return a clear "nothing matched" message rather than misleading output, and answers are reconciled against what is displayed.