Sluice Docs

Sluice Documentation

Sluice provides guardrails for AI-generated email. It sits between your AI agents and your customers, checking every outbound email against configurable safety guardrails before it gets delivered.

However your agent sends email — REST API or SMTP — and whatever provider you use, Sluice intercepts, inspects, and forwards only what passes your guardrails.

How it works

Your AI Agent  ──REST API──>  Sluice Guardrails  ──>  Customer Inbox
Your AI Agent  ──SMTP──────>  Sluice Guardrails  ──>  Customer Inbox
                                    |
                                    |  flagged?
                                    v
                               Review Dashboard
                               (approve / edit / reject)
  1. Send email through Sluice — via the REST API (programmatic control) or SMTP (drop-in, no code changes)
  2. Guardrails analyze every email for PII leaks, tone issues, policy violations, prompt injection, and more
  3. Safe emails forward automatically to your customers via your configured outbound provider
  4. Flagged emails are held for human review in a simple dashboard

Why Sluice?

AI agents are increasingly sending emails on behalf of businesses — customer support replies, follow-ups, notifications, outreach. But without guardrails, a single hallucinated price, leaked SSN, or unprofessional tone can damage trust and create liability.

Sluice gives you confidence that every AI-generated email meets your standards before it reaches a customer.

  • Two integration paths — REST API for full programmatic control, or SMTP for zero-code-change drop-in setup.
  • 10 configurable guardrails — From PII detection to hallucination checking to rate limiting.
  • Human review dashboard — A clean two-pane interface for reviewing, editing, and approving flagged emails.
  • Works with your email provider — Sluice forwards approved emails through Resend, SendGrid, Postmark, Gmail, Outlook, or any SMTP server.

On this page