Skip to main content

Persado MCP User Guide: Claude

Learn how to score and improve content in Claude using Persado's MCP.

A
Written by Amy Blakemore

The Persado MCP connects Claude to Persado’s machine learning models, trained on real campaign results. This guide walks you through how to score content and improve content directly in Claude.

In order to use the capabilities below, you must first install the Persado MCP Connector in Claude; this should only take a few minutes! See here for instructions.

How to Improve Content in Claude

  1. Log into your Claude instance as usual.
    ​

  2. To improve upon existing content, simply type in or attach your copy into a new project. You can also use other existing connectors that your team leverages to import the content (e.g., from Figma or Adobe Workfront). You don’t need to explicitly ask Claude to leverage the Persado MCP; it will pull it in automatically during the tool call. You’ll see the Persado logo while it’s running commands.
    ​

  3. The first time you initiate a request to improve content, you will need to Always Allow to move forward. You will only have to do this once.
    ​

  4. Next, you’ll see your original message (i.e., your control message) and top-performing alternatives. Every message, including your control, will be accompanied by the following information:

    1. Performance Score: How well content is predicted to perform. The higher the score, the more likely your copy is to outperform other messages in your industry. All scoring is based on Persado’s 10 years of engagement data. To learn more about Performance Scores, click here.

    2. Brand Fit: A 1 to 5 scale rating of how well a Variant aligns with the model’s understanding of your brand (1 being the lowest, 5 being the highest).

    3. Compliance Risk Level: A way to assess compliance risk at a glance. Persado quantifies risk at three levels—low, medium, and high—and always includes an explanation for the risk level and how it could be reduced.

    4. Why it Scores: An explanation behind the scores so you can understand the language elements at play. This section will also note the language tags used in the copy to show you at a more granular level what is working.
      ​

  5. Review the copy and explanations. The MCP will designate a “top pick” for you based on the content that scores the highest. It will also note a short explanation for your reference, including any pointers on how to reduce compliance risk. Note that you can ask for more options as many times as you like, and provide feedback for what you’d like adjusted the same way you would with Claude as usual!
    ​

  6. From here, we recommend you follow your business as usual process. This might include:

    1. Asking Claude to generate a document of the outputs so you can circulate with your internal team for review—such as your legal team.

    2. Copy and pasting your preferred message into your ESP/Marketing Platform and deploying as usual.

    3. If you have an agentic tech stack, using connections you already have configured in Claude to push content to your workflow or content libraries (e.g. Figma or Jira MCP connectors).

How to Score Content with Claude

If you’re looking for guidance when creating new content, using the MCP’s scoring capability is beneficial for getting a gut check. This might be useful, for example, if you’re populating marketing language into a library and want a performance check before you upload. Or, if you have connectors to your ESP/DAM/creative repository, try scoring those assets through Persado to identify which to focus on for optimization!

  1. Log into your Claude instance as usual.
    ​

  2. To score existing content, simply type in or attach your copy into a new project, or use existing connectors to your ESP/DAM/creative repository.
    ​

  3. The first time you initiate a request to score content, you will need to ‘Always Allow’ to move forward. You will only have to do this once.
    ​

  4. The same with improving content, you’ll see a visual breakdown of your copy and its related scores for performance, brand fit, and compliance. Review all of the data and share with your team as desired, especially if there are flags around compliance that you would like to resolve before deploying to your audience.

FAQs

What marketing channels does the Persado MCP server support?

The Persado MCP supports the following channels:

  • Full Email

  • Email Subject and Body

  • Call Center Script

  • Out-of-Home Ad

  • POS

  • Flash SMS

  • Radio Script

  • Gmail Sponsored Promotion

  • Snapchat

  • TikTok

  • Automated Call Prompt

  • Inline Ad

  • WhatsApp

  • MMS

  • Desktop Notification

  • App Store Page

  • Customer Support Chat

  • Push Notification and In-App Body

  • Web Page

  • In-App Body

  • Web Banner

  • Push Notification

  • Mobile App

  • Paid Search (Google Ads)

  • Facebook Ad

  • Email Banner

  • Display Banner

  • Direct Mail

  • LinkedIn Ad

  • Sms/Ems

  • Billboard

  • Flyer/Brochure.

How quickly should the Persado MCP server respond?

Response rates, like with any LLM, will depend on Claude’s server capacity, but we estimate that the Persado response will take under 60 seconds. Claude will then typically take 1 to 2 minutes to visualize the response, though it can take up to 5 minutes at the high end.

If there are delays in the response, we highly recommend you check Claude’s operational status here, and sign up for email alerts so you can be notified real time of issues: https://status.claude.com/

What is the best way to export content and scores to share with my team?

There are a few ways to collaborate with your team:

  • Send content and scores straight to Slack or your preferred project management tool using existing Claude connectors.

  • Copy content and scores straight from the LLM and paste as desired.

  • Ask Claude to export Persado’s outputs in the format you are most accustomed to already working in with your team. This might be a Word document or an Excel sheet.

Why doesn’t the Persado MCP generate low compliance risk alternatives as a default?

Every team’s tolerance for risk is different, and this scale is simply a guide to calculate, at a glance, how close the creative is to being compliant based on Persado’s general framework. While ideally content would have zero compliance risks, this is often challenging for an AI to validate without the complete context of a communication (disclosures, full user journeys, targeting methods), so keep a pulse on your team’s own needs. Your compliance team should be able to weigh in on the risks to monitor closely—not just the score. Context matters!

Do edits and feedback I provide in Claude flow back to Persado?

Not currently, but this is on the roadmap!

Did this answer your question?