Sender MCP Server

Discover practical AI use cases to boost your marketing efficiency.

This guide explains how to connect your Sender account to an AI assistant using the Sender MCP server, what the assistant can do once connected, and the safeguards that keep it from sending anything on its own.

Where to Find This Feature

There is nothing to enable in the Sender dashboard. You connect from your AI assistant’s side using the Sender MCP server URL:
https://mcp.sender.net/mcp
Authentication uses OAuth, so you log in to Sender in a browser window and approve access. There is no API key to copy or paste, and you can revoke access at any time.

What You Need Before Connecting

An MCP-compatible client. Claude, ChatGPT, and Cursor all support remote MCP servers. Availability depends on your plan with those products, so check their documentation if the option isn’t visible.

A Sender user role with the right permissions. Every tool checks a permission against the role of the user who authorized the connection. If your role can’t edit campaigns in the dashboard, the assistant can’t edit them either — it receives a permission error and explains why.

A verified sending domain, for anything that will eventually send. Campaign drafts and workflow email steps both need a from_email on a verified domain. Ask your assistant to list your sending domains and it reports verification, SPF, DKIM, and DMARC status for each one.

How the MCP Server Works

MCP (Model Context Protocol) is an open standard for connecting AI models to outside tools and data. Connecting Sender’s MCP server gives your assistant a set of tools it can call on your behalf — reading your subscribers, segments, campaigns, and automations, and creating things in your account.

Everything the server creates arrives as a draft. Campaigns are never sent through this connection at all. Workflows are created inactive, and going live is a separate step that runs the same pre-flight checklist as the dashboard.

Connecting Sender to Claude

Here are steps to take in order to Connect Sender to Claude.

1. Go to Customize → Connectors in Claude.
2. Click +, then Add custom connector.
3. Enter Sender as the name and https://mcp.sender.net/mcp as the URL, then click Add.
4. Select the Sender connector and click Connect. Log in to Sender and approve the permissions.
5. Start a conversation and ask for something that needs your account data. “List my segments” is a quick test.

Here are the steps in action:

Note: On Claude Team and Enterprise plans, a Claude organization Owner has to add the connector at the organization level before members can connect their own Sender accounts.

Once connected, the Sender connector is available in Claude on web, desktop, and mobile.

Connecting Sender to ChatGPT

Adding a custom MCP server in ChatGPT requires developer mode.

Here are steps to take in order to Connect Sender to ChatGPT:

1. Open Settings → Plugins, click “Developer mode” and turn on the Developer mode toggle.
2. Return to Plugins and click Browse plugins.
3. Click “+” sign.
4. Enter Sender as the name, https://mcp.sender.net/mcp as the Connection Server URL, and set authentication to OAuth.
5. Complete the OAuth flow — log in to Sender and approve the permissions.
6. Select the Sender app in the composer before asking a question that needs your account data.

Here are the steps in action:

Connecting Sender to Cursor

1. Open Cursor Settings → MCP → New MCP Server, or edit your MCP configuration file directly.
2. Add a server entry:

json

{
  "mcpServers": {
    "sender": {
      "url": "https://mcp.sender.net/mcp"
    }
  }
}

3. Save, then complete the OAuth flow when Cursor prompts you.
4. Start a chat and confirm the Sender tools appear.

Using Multiple Sender Accounts

Each connection is authorized against one Sender account. To work across several, add a separate connector entry per account, give each a distinct name (for example sender-brand-a and sender-brand-b), and authorize each one with the login for that account.

Enable only the connector you intend to work in for a given conversation. That way an assistant can’t read one account and write to another by mistake.

What You Can Do With It

Here are some examples of prompts you can give:

• “Act as an experienced email marketer. Review my segments and tell me which ones are broken, duplicated, or misnamed.”
• “Which of my automations overlap? Show me what a new free user actually receives in their first two weeks.”
• “Compare my three upsell workflows — sends, opens, clicks, and unsubscribes side by side.”
• “Check the emails in my onboarding flow for broken links, missing preheaders, and wrong image alt text.”
• “Draft a short welcome sequence for new signups and build it as a draft workflow.”

Auditing and Building Segments

Your assistant reads every condition in a segment, not just its name, which is how it catches filters that quietly stopped working.

Fixed dates are the usual culprit. A segment named “30-day most engaged” built with a specific date two years ago now means “opened anything in the last two years” — and if it holds a third of your list, anyone reaching for it expects a tight audience and mails the database instead.

The same read surfaces duplicate segments with identical conditions, names that promise a filter the segment doesn’t apply, and lists holding unsubscribed records.

You can then describe replacements in plain language — “signed up more than 12 months ago, zero opens and zero clicks in the last 365 days” — and the assistant writes them in Sender’s condition vocabulary and creates them. Every segment comes back with a plain-English readback of its conditions so you can check the logic before you use it.

Reviewing Your Automations

Ask what a subscriber actually receives and you get the merged picture across every active workflow, which is harder to see in the dashboard than it sounds.

Two automations sharing a trigger event means everyone who enters one enters the other. A third gated on an event that fires inside the second makes its audience a subset of that one. Stack the delays and you can find five emails landing in nine days, two of them on the same morning.

Your assistant can lay that out on a timeline, then pull the performance report for each workflow to show what the overlap costs in unsubscribes. It also reads the structure itself: which steps block activation and why, and any step no path can reach.

Checking Email Content

Your assistant can read the body of a campaign or a workflow email step, and it notices details that are easy to miss in a visual editor:

• Social icons carrying the wrong alt text — a YouTube icon labeled alt="Twitter", LinkedIn labeled alt="Instagram"
• Placeholder alt text such as alt="Image description" left on a main image
• Hero images linking somewhere that contradicts the email, like a registration link in a message sent only to people who already registered
• Legacy {$firstname} tokens, which reach subscribers as literal text. Sender uses Liquid, so personalization is {{ firstname }}. Every content write scans for the old syntax and warns you
• Missing preheaders, and permission reminders too vague to rely on if someone complains

Building a Workflow

You can go from a description to a working draft in one conversation. Ask for a welcome sequence and your assistant resolves the group and event IDs it needs, then submits the whole graph — trigger, delays, conditions with their branches, splits — in a single step.

It comes back as a draft, with a plain-English rendering of the branching so you can confirm the logic reads the way you meant, plus a list of anything still blocking activation. From there the assistant fills in the email bodies, runs a dry-run validation that reports every remaining problem at once, and activates only when you tell it to.

Editing later is non-destructive. Inserting steps splices existing ones underneath rather than replacing them, and deleting a step won’t silently drop subscribers waiting on it.

Available Tools

Your MCP client discovers the full tool list and each tool’s inputs automatically. The complete reference, including conventions and multi-tool recipes, is in the Sender API documentation.

Access legend: R read-only · W writes · D destructive (requires an explicit acknowledgment flag)

Account and Discovery

Tool	Access	What It Does
get_account_details	R	Account name, timezone, country, status, plan, and calling user
list_domains	R	Sending domains with verified, SPF, DKIM, and DMARC status
list_phone_numbers	R	SMS sender numbers — required before any SMS step
list_stores	R	Connected stores — required for store-driven triggers and conditions
get_event_types	R	Custom and built-in event types, with the IDs event triggers use
get_events	R	Event instances for one type, newest first

Subscribers and Groups

Tool	Access	What It Does
list_subscribers / get_subscriber	R	Browse subscribers, or fetch one by ID, email, or phone
create_subscriber / update_subscriber	W	Create or update profile fields, groups, and channel statuses
subscribe_subscribers / unsubscribe_subscribers	W	Reactivate or unsubscribe people per channel
list_groups / get_group	R	Subscriber groups

Segments

Tool	Access	What It Does
list_segments / get_segment	R	Browse segments; the single fetch returns conditions plus a plain-English readback
create_segment / update_segment	W	Build segments from conditions; updates replace all conditions at once

Email Campaigns

Tool	Access	What It Does
list_email_campaigns / get_email_campaign	R	Browse campaigns; the single fetch includes audience, reports, and the body
get_email_campaign_report	R	Performance report for a sent campaign
create_email_campaign	W	Draft campaign with sender details, audience, and optional body. Requires a verified sending domain. Never sends
update_email_campaign_content	W	Replace a draft campaign’s body

Workflows

Tool	Access	What It Does
list_workflows	R	Browse automations with light engagement stats
get_workflow	R	The full graph, a plain-English rendering of the steps, and what blocks activation
get_workflow_report	R	Performance report, with revenue where your plan includes it
get_workflow_step_content	R	One step’s sender details and body, or an SMS step’s text as it will send
validate_workflow	R	Dry run — structural checks before creation, or the full activation checklist
create_workflow	W	The whole graph in one step: title, trigger, and nested steps. Always a draft
rename_workflow	W	Title only
update_workflow_step	W	Reconfigure one step by ID. Step type can’t be changed
add_workflow_steps	W	Insert steps anywhere; existing steps re-attach underneath
delete_workflow_step	D	Remove a step or a branch. Won’t drop waiting subscribers without an explicit flag
update_workflow_email_content	W	An email step’s body
update_workflow_sms_content	W	An SMS step’s message and sending number, up to 1,600 characters
activate_workflow	W	Runs the full pre-flight checklist and takes the workflow live
deactivate_workflow	W	Active back to draft. Nothing is lost, and it’s required before any edit

Image Library

Tool	Access	What It Does
list_image_categories	R	Image folders
upload_image_from_file / upload_image_from_url	W	Add images for use in email content

What the MCP Server Will Not Do

Worth reading before you give an assistant access to your account.

Creation is always a draft. Campaigns are never sent through this connection. Workflows go live only through one deliberate activation step, which runs the same pre-flight checklist as the dashboard. If anything blocks activation, nothing activates and you get every problem at once rather than one at a time.

Active workflows can’t be edited here. Deactivate first. All subscriber progress is preserved, and everyone resumes where they left off when you reactivate.

Destructive actions name what they affect. Dropping subscribers waiting on a step, or discarding a drag-and-drop design, each require an explicit flag rather than happening quietly.

Workflow creation is rate-limited to 50 per account per day, as a backstop against a runaway assistant.

Drag-and-drop designs are protected. Emails built in the block builder are read-only through this connection unless you explicitly allow an overwrite, which permanently discards the design.

What Happens Next

Once the connector is authorized, your assistant sees your account immediately — no sync or import step. Anything it creates appears in your Sender dashboard right away as a draft, where you can review it, edit it in the normal editors, and send or activate it yourself.

Common Issues

Can’t connect → Check the URL is exactly https://mcp.sender.net/mcp. Your assistant connects from its own cloud infrastructure rather than from your computer, so a VPN or local firewall isn’t usually the cause, though a client-side proxy can be.

Authentication fails → Confirm you completed the OAuth flow in the browser window and approved the permissions. If your organization applies plan-level controls to your AI assistant, an administrator may need to add the connector before you can authorize it.

Tools are missing → Restart your client and check the Sender connector is enabled for that specific conversation or workspace. Some clients require you to select it in the composer each time.

A tool returns a permission error → The connection carries your Sender role. If you can’t perform the action in the dashboard, you can’t perform it here either. Ask an account owner to connect instead, or to grant your role the permission.

A workflow won’t activate → Ask your assistant to validate it and you’ll get every blocking problem in one response. The most common cause is an email step whose from_email isn’t on a verified sending domain.

Personalization shows up as literal text → Sender uses Liquid: {{ firstname }}, not {$firstname}. Content written with the old syntax is stored exactly as passed and reaches subscribers unrendered.

FAQs

Can an AI assistant send emails to my subscribers? No. Campaigns created through the MCP server are always drafts, and there is no tool that sends one. Workflows can go live, but only through a separate activation step you ask for explicitly, and only after passing the same pre-flight checks the dashboard runs.

Do I need a paid Sender plan to use this? The MCP server follows your account’s existing permissions and plan features. If a feature isn’t included in your plan, it isn’t available here either.

Does my assistant see my whole account? It can read the data your Sender user role gives you access to. Tools check permissions individually, so a limited role produces a limited view.

Can I disconnect it? Yes. Remove the connector in your AI assistant. The connection stops immediately and nothing created before then is affected.

Will it change things without asking? Your assistant proposes actions and most clients ask you to approve tool calls that write data. Review write and delete steps before approving them, the same way you would any integration with account access.

Can I use it with an assistant other than Claude, ChatGPT, or Cursor? Any client that supports remote MCP servers with OAuth should work. Those three are the ones we’ve documented.

---

If you got stuck on a specific task or can’t find a way to execute a particular job, contact our support team via LiveChat or support@sender.net — we’re here to help 24/7.