What is Sluice?

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

Think of it like a safety net — it filters outgoing emails for PII leaks, hallucinations, tone issues, policy violations, and more.

The problem

AI agents are sending more and more emails on behalf of businesses: customer support replies, sales follow-ups, appointment confirmations, onboarding sequences. But AI-generated emails can go wrong in ways that are hard to predict:

• An agent hallucinating a price or feature that doesn't exist
• An agent leaking a customer's SSN or credit card number in a reply
• An agent adopting an aggressive tone when handling a complaint
• A prompt injection attack causing an agent to send unauthorized content
• A malfunctioning agent loop sending thousands of duplicate emails

One bad email can destroy customer trust, trigger compliance violations, or cause reputational damage. Manual review of every email doesn't scale. You need automated guardrails with human oversight for edge cases.

The solution

Sluice works with your existing email stack. Two integration paths:

• REST API — Send emails programmatically with guardrails built in. No SDK needed, just HTTP calls. Ideal for new integrations and AI agent frameworks.
• SMTP — Point your agent's SMTP configuration at Sluice. No code changes beyond updating host/port/credentials. Ideal for existing setups.

Key features

• 10 built-in guardrails — PII detection, tone analysis, content policy, hallucination checking, prompt injection defense, rate limiting, recipient rules, attachment scanning, duplicate detection, and compliance checks
• REST API and SMTP ingestion — Use the REST API for programmatic control, or SMTP for drop-in integration with zero code changes
• Human review dashboard — A clean two-pane interface where reviewers can approve, edit, or reject flagged emails
• Tuning mode — Hold all emails for review while you dial in your guardrail settings, then flip the switch to auto-forward safe emails
• Bring your own email provider — Sluice forwards approved emails through your existing provider: Resend, SendGrid, Postmark, Gmail, Outlook, or any SMTP server
• Knowledge base — Upload your product docs, pricing, and policies as a source of truth for hallucination detection
• Audit log — Full trail of every email, guardrail result, and reviewer action for compliance reporting

Who is Sluice for?

Sluice is for any team that has AI agents sending customer-facing emails and wants to make sure those emails are safe, accurate, and on-brand before they go out.

Common use cases:

• Customer support teams using AI agents to draft and send replies (via n8n, OpenAI Agents, LangChain, CrewAI, or custom builds)
• Sales teams using AI to automate outreach and follow-ups
• Operations teams with AI agents sending notifications, confirmations, and updates
• Regulated industries (healthcare, finance, legal) that need compliance guardrails on all outbound communication
• Agencies and consultancies managing AI agents on behalf of multiple clients

How it works in 30 seconds

AI Agent  ──REST API / SMTP──>  Sluice  ──guardrails──>  Customer Inbox
                                  |
                                  |  flagged?
                                  v
                             Review Queue

1. Your AI agent sends an email through Sluice — via the REST API or SMTP
2. Sluice receives the email and runs it through your configured guardrails
3. If all guardrails pass, the email is forwarded to the recipient automatically
4. If any guardrail flags a concern, the email is held for human review
5. A reviewer can approve the email as-is, edit it and approve, or reject it

The recipient never knows Sluice exists. Approved emails are sent from your own email address, through your own email provider.