Skip to main content

Documentation Index

Fetch the complete documentation index at: https://notikaai.mintlify.app/llms.txt

Use this file to discover all available pages before exploring further.

An AI email agent turns an intent into an email workflow. Instead of writing every trigger, query, template, wait, and condition by hand, you describe the outcome you want. The agent uses your app context to draft the workflow and lets you preview it before it goes live.

​
What makes it different from AI copywriting?

AI copywriting helps write words. An AI email agent helps build the whole workflow:
  • who should receive the email
  • when it should start
  • what data should be checked
  • whether to wait
  • what should happen after the wait
  • what the email should say
  • what template and brand treatment to use
  • how to preview and publish safely
The copy matters, but the workflow is the product behavior. That is why Dreamlit is end-to-end: AI can help across the system, not only with the words inside the email.

​
Example

You can ask:
Welcome new users, wait 3 days, then send a reminder only if they have not created their first project.
A useful agent needs to turn that into:
  1. a trigger for new users
  2. a wait step
  3. a query against current user or project data
  4. a conditional send
  5. a branded email template
  6. a preview with real rows

​
Where MCP fits

Dreamlit exposes an MCP server so AI tools like Codex, Cursor, Claude, and other clients can inspect your Dreamlit workspace and help create workflows. That means an agent can help you from the place you already work:
  • your codebase
  • your database schema
  • your Dreamlit workflow builder
  • your docs and product context

Dreamlit MCP server

Connect an AI client to Dreamlit MCP.

​
When should you use one?

Use an AI email agent when you know the product outcome but do not want to manually assemble the workflow infrastructure. Good prompts look like:
  • “Send a usage warning every morning to customers above 80 percent of their plan.”
  • “Remind invited teammates after 48 hours if they have not accepted.”
  • “Send a weekly digest to admins with new signups and failed payments.”
  • “Send Supabase Auth confirmation emails with our brand voice.”
  • “Send a product update to active customers who use the affected feature.”

​
What should stay human-reviewed?

Always review before publishing:
  • audience rules
  • unsubscribe behavior
  • timing
  • sample rows
  • email copy
  • sender identity
The agent can draft. You still decide what should happen to real users.