Workflow Examples
End-to-end workflow examples showing how to chain Keplars MCP tools to complete real tasks through natural language.
These workflows show how to combine multiple MCP tools in a single conversation to complete a real task. Copy the prompts into your AI client and adapt them to your own data.
IDs matter
Most tools reference audiences, templates, and campaigns by ID. If you do not know an ID, ask the AI to list them first. For example: "List my Keplars audiences" before "Create a campaign targeting audience aud-xyz".
Workflow 1: Run a newsletter campaign from scratch
This workflow takes you from zero to a sent campaign: pick a template, pick an audience, create the campaign, confirm, and send.
Step 1: Discover what you have
"List my Keplars templates and audiences"
The AI calls list_templates and list_audiences in sequence and returns the IDs and names of both. Note the template ID and audience ID you want to use.
Step 2: Create the campaign
"Create a campaign called 'September Newsletter' targeting audience aud-abc using template tmpl-xyz with subject 'What is new this month'"
The AI calls create_campaign and returns the new campaign ID.
Step 3: Send with confirmation
"Send campaign camp-123"
The AI calls send_campaign. Before dispatching, it fetches the campaign details and shows you the audience size:
About to send 'September Newsletter' to 4,200 contacts. Confirm? (yes / no)Reply yes to dispatch. The campaign enters the queue immediately.
Step 4: Check delivery
"What are the stats for campaign camp-123?"
The AI calls get_campaign_stats and returns the sent count, failed count, open count, and current status. Run this a few minutes after sending for the first numbers.
Workflow 2: Add a new user and send a welcome email
Use this when a new user signs up outside your main app and you want to add them to your mailing list and trigger a welcome email in the same turn.
Single prompt approach
"Add [email protected] to audience aud-newsletter with first name Sophie and tag 'trial'. Then send her the welcome template tmpl-welcome with variables first_name=Sophie and plan=Trial from [email protected]"
The AI runs two tool calls back to back:
add_contactto create the contact in the audiencesend_templated_emailto send the welcome email with her name and plan injected
No confirmation step is needed because this is a single transactional send, not a bulk operation.
Verify the contact was added
"How many contacts are in audience aud-newsletter now?"
The AI calls get_audience and returns the updated count.
Workflow 3: Clean up an audience
Use this to inspect an audience and remove a specific contact, for example when handling an unsubscribe manually.
Step 1: Check the audience
"Get the details of audience aud-newsletter"
The AI calls get_audience and returns the name, contact count, and creation date.
Step 2: Remove a contact
"Remove contact cid-88f3a2 from audience aud-newsletter"
The AI calls remove_contact. The contact is removed immediately.
Contact IDs are required
remove_contact requires the contact ID, not the email address. If you only have the email, look up the contact ID in the Keplars dashboard before asking the AI to remove it.
Step 3: Confirm the count dropped
"How many contacts are in audience aud-newsletter now?"
The AI calls get_audience again and returns the updated count.
Workflow 4: Send a quick transactional email
The simplest workflow. No setup required, just a single prompt.
"Send a transactional email to [email protected] from [email protected] with subject 'Your invoice is ready' and body 'Hi, your invoice for August is attached at invoice.yourdomain.com/inv-9900'"
The AI calls send_transactional_email and the email enters the instant queue. Expected delivery is under 5 seconds.
Check your verified domains if the send fails
"Which domains have I verified in Keplars?"
The AI calls list_verified_domains. If the domain in your from_email is not in the verified list, the send will be rejected. Verify the domain in the Keplars dashboard first.
Tips for writing good prompts
Be explicit about IDs when you know them
"Create a campaign targeting audience aud-abc" is faster than "Create a campaign targeting my newsletter list" because the AI does not need to list audiences first.
Chain operations in one message
You can ask the AI to list, pick, and act in a single prompt. It will run each tool in order and pause for confirmation only when a bulk operation is involved.
Use the confirmation step as a safety check
For campaign sends and bulk emails, the AI always asks for confirmation with the audience size shown. Read it before replying "yes". This is the last point to catch a wrong audience or wrong template.
Ask for stats after every send
"What are the stats for campaign camp-xyz?" a few minutes after sending gives you early delivery numbers. Re-run the same prompt after an hour for open rates.