MCP server use cases
Each of these is work a marketer normally spreads across several Accelerator screens, done instead as one conversation. The examples show the goal, a sample prompt, what happens, where the time savings comes from, and where the assistant pauses to confirm. Pick your Accelerator version to see which of these you can use.
1. Build a launch-ready campaign from a prompt​
✓ 26.3.2+"Create a campaign 'Spring Sale 2026' from the 'Seasonal Promo' template, attach my 'Active Buyers' audience, and run the launch checklist."
The assistant creates the campaign from the template, attaches the audience, runs the pre-send launch checklist, and hands you a link to the campaign, ready to launch in Accelerator. What normally means visiting the campaign builder, the audience picker, and the checklist in turn becomes one prompt and one review.
2. Create and iterate an audience​
✓ 26.3.2+"Build an audience of subscribers who purchased in the last 90 days but haven't opened an email in 30. How many rows is that?"
The assistant creates the audience and reads back its detail, including the row count, and then you iterate, for example "exclude anyone who unsubscribed", checking the count as the definition changes. Each targeting revision takes a sentence rather than a round of editing SQL by hand and re-running it. Validate targeting before attaching an audience to a campaign.
3. Test a send safely​
✓ 26.3.2+"Make a test version of this audience and attach it to the campaign so I can try it."
The assistant creates a test audience in which every recipient routes to your configured test inbox and attaches it to the campaign, so the launch you run from Accelerator exercises the real send path without reaching real subscribers. Use this before any production send.
4. Draft copy that fits the channel​
✓ 26.3.2+"Three subject-line options for a win-back email, then tighten the one I choose so it won't get truncated."
The assistant writes the options itself and checks them against the platform's real limits, so review starts immediately instead of after a drafting cycle. Because it knows which limits actually truncate a message and which are advisory, you get copy that fits the channel rather than copy you discover is too long at proof time.
5. Make a template multichannel​
✓ 26.3.2+"Take the 'Spring Sale' template and add an SMS version and push-notification content for iOS and Android."
The assistant writes channel-appropriate copy and sets the SMS body and per-provider push content on the template, so the same template can send on email, SMS, and push. One conversation replaces re-authoring the same message in each channel's editor. Removing a channel's content later works the same way.
6. Deliver an audience to ad platforms​
✓ 26.3.2+"Set up an external campaign that syncs my 'High Propensity' audience to our Meta Ads and Google Ads destinations."
The assistant creates the external campaign and attaches the audience and destinations, collapsing setup that spans the audience, destination, and campaign screens into one pass. Launching an external campaign happens in Accelerator, not through the assistant; the assistant hands you a link to the campaign it configured, ready to launch.
7. Pre-launch QA and go/no-go​
✓ 26.3.2+"Run the launch checklist on the Spring Sale campaign and give me a go/no-go."
Runs the pre-send checklist: template attached, audience attached and non-empty, sender identity resolvable, and a sane schedule. Each item comes back as pass, warning, or fail, ending with a clear recommendation, so you walk into the launch screen already knowing the answer.
8. Check whether a send went out​
✓ 26.3.2+"Did the Spring Sale send go out? What's the job status?"
The assistant reads the send job's status and reports where it stands, with a link to the campaign in Accelerator for the full picture. Instead of hunting for the campaign in the UI to confirm a launch, you ask in a sentence.
9. Review how a campaign performed​
✓ 26.3.2.1+"How did the Weekly Newsletter perform this year? Show me its open rate by month, and tell me whether engagement is going up or down."
The assistant reads the campaign's results across its sends and its results over time, then summarizes them as a trend rather than a table of numbers. If a metric isn't available, the assistant says so rather than reporting it as zero. Instead of opening the campaign's analytics and setting the date range and chart yourself, you ask the question you actually have. Campaign analytics require Accelerator 26.3.2.1 or later.
10. Find where a send underperformed​
✓ 26.3.2.1+"How did yesterday's send of the Fall Sale perform? Which email domains had the lowest delivery rate, and what share of opens came from mobile devices?"
The assistant reads the send's results and breaks them down by email domain, device, or email client, so you can see whether a drop comes from one mailbox provider or across the board. What normally means switching between several breakdown views for one job becomes a single question, with a link to the campaign in Accelerator for the full picture. Single-send analytics require Accelerator 26.3.2.1 or later.
Tips for good results​
- Be specific about audiences. Check the row count before attaching an audience to a campaign.
- Use a test audience first. Have the assistant attach a test audience that routes to your inbox before any real launch.
- Ask it to explain, not just act. For example, "explain what this will do before you change anything."
- Confirm deliberately. Read the summary before approving a delete.
- When unsure, ask read-only. For example, "just show me, don't change anything."