FAQ
General
What is Sluice?
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 — via REST API or SMTP. Think of it like a firewall, but for outgoing emails.
Do I need to change my code to use Sluice?
It depends on your integration path. With SMTP, you just update your host, port, and credentials — no code changes. With the REST API, you send emails via HTTP calls instead of through your current provider. Either way, setup takes about 5 minutes.
What AI agent platforms does Sluice work with?
Any platform that can send email via SMTP or make HTTP calls. This includes OpenAI Agents, LangChain, CrewAI, AutoGen, Semantic Kernel, n8n, Make, Zapier, Power Automate, and any custom application written in any programming language. See AI agent examples for copy-paste integration guides.
Does the recipient know Sluice is involved?
No. Approved emails are sent from your own email address through your own email provider. Sluice is completely transparent to the recipient.
Setup
How long does setup take?
About 5 minutes. Sign up, configure your outbound email provider, and connect your agent via the REST API or SMTP.
Can I use Sluice with multiple AI agents?
Yes. All agents within your organization share the same SMTP credentials and guardrail configuration. Every email from every agent goes through the same guardrails.
Do I need to configure my outbound email provider?
Yes. Sluice forwards approved emails through your own provider (Resend, SendGrid, Postmark, Gmail, Outlook, or any SMTP server). You configure your provider credentials in Settings > Outbound. See Outbound providers.
Guardrails
Which guardrails should I enable?
Start with the defaults — PII Detection, Tone Analysis, Content Policy, and Prompt Injection are enabled out of the box and provide strong baseline protection. Enable additional guardrails (Rate Limiting, Recipient Rules, etc.) based on your specific needs. See the Guardrails overview.
Can I customize the guardrail rules?
Yes. Most guardrails have configurable settings — thresholds, custom definitions, allowed/blocked lists, and more. See individual guardrail pages for configuration options.
What happens if a guardrail incorrectly flags an email?
The email is held for review in your dashboard. You can approve it as-is or edit it before approving. To reduce future false positives, adjust the guardrail settings (e.g., whitelist specific PII types, update your tone definition, or adjust confidence thresholds).
Can I add my own custom guardrails?
The 10 built-in guardrails cover the most common needs, and several of them (Tone Analysis, Content Policy, Compliance) support custom definitions that let you tailor them to your specific requirements. If you have needs beyond the built-in guardrails, contact us.
Security
How are my SMTP credentials stored?
Your inbound SMTP password is bcrypt-hashed before storage — Sluice never stores it in plaintext. Your outbound SMTP passwords and OAuth2 tokens are encrypted at rest using AES-256-GCM and only decrypted at send time.
Is email content encrypted?
Yes. Email bodies are encrypted at rest. They are decrypted only when needed for guardrail analysis or when displayed in the review dashboard.
Who can access my emails?
Only members of your organization who have been invited via the dashboard. Admins and reviewers can see emails in the review queue. Each organization's data is completely isolated.
Email delivery
Does Sluice affect my email deliverability?
No. Sluice forwards approved emails through your own email provider and sending domain. Your deliverability, reputation, and authentication (SPF, DKIM, DMARC) are unchanged.
What happens if Sluice is unavailable?
If Sluice is unreachable, your request will fail — with an HTTP error for REST API calls, or a standard SMTP error for SMTP connections. Most HTTP clients and SMTP libraries have built-in retry logic that will attempt to resend when Sluice is available again. No emails are lost.
How fast is the guardrail analysis?
Most emails are analyzed within a few seconds. The exact time depends on which guardrails are enabled and the complexity of the email content. Emails that pass all guardrails are forwarded immediately when tuning mode is off.