Beta: Paramark Model Context Protocol (MCP)

In November 2026, we're launching Paramark's MCP in beta (available with the Advanced and Enterprise plans).

Click here for more information on Paramark plans and pricing

What is Paramark's MCP and why we built it?

Here's how it works

The Paramark connector brings your Paramark data into the AI tools you already use, including ChatGPT and Claude, so you can put it to work elsewhere: building presentations, drafting messages, or combining it with other data and workflows.

The connector is built for accessing your data, not for interpreting it. For answers, analysis and recommendations about your marketing measurement, use Leo in Paramark. Leo is built on your models and our methodology, while a general-purpose assistant doesn't have that context and tends to give more generic, less informed answers.

Your assistant can access:

• Marketing Mix Modeling (MMM) results: what each channel contributed, cost per incremental outcome, and incremental return on ad spend for any published report, by week, month or quarter.

• Live MMM: how channels are trending since your last published report, through the most recent data.

• Experiments: what a geo test was designed to detect, and the lift it measured.

• The data behind the models: spend and impressions by channel, and your target metrics over time.

In the current version the connector is read-only. It cannot change a model, a report, a plan or anything else in Paramark.

Why we built it

Much of the work that follows a measurement result happens outside of Paramark: in slide decks, team updates, planning spreadsheets and the AI tools your team already uses every day. We built the MCP so your Paramark data can be easily accessed in those cases, without exporting files or copying numbers by hand.

On the other hand, Leo still exists as the place to interpret that data. Leo has the context of your models and our methodology, so we do not recommend relying on a more generic assistant for answers about what your results mean. The MCP gets your data where you need it, and Leo helps you understand it.

How to use Paramark's MCP

Before you start

You need three things:

• A Paramark login: You sign in with the same email you use for the Paramark dashboard. There are no extra passwords or API keys to manage.

• An Advanced or Enterprise Plan: Only these two plans include access to the Paramark MCP server. Message your growth advisor in Slack for more information.

• ChatGPT or Claude: Add Paramark from the directory, or as a custom server. Other tools work too, as long as they can connect to an MCP server.

Once connected, the assistant sees exactly what you see in the dashboard: the reports, experiments and data in the Paramark account you signed in to.

Add Paramark to your assistant

Whichever route you take, the last step is the same: a Paramark sign-in page opens, you sign in, and you are back in the chat. If your login belongs to more than one Paramark account, you are asked to pick one, and the connector reads that account until you reconnect. Your sign-in stays valid for a while. When it expires, the assistant asks you to reconnect.

In ChatGPT

• Open chatgpt.com/plugins and search for Paramark.

• Select Add, then sign in to Paramark when prompted.

• Ask a question. ChatGPT picks up the connector on its own, or you can type @Paramark to point it there.

In Claude

• Open Settings, then Connectors, and browse the directory for Paramark.

• Select Connect, then sign in to Paramark when prompted.

• In a new chat, check that Paramark is enabled under the tools menu, and ask away.

As a custom MCP server

Use this in a tool that is not listed above, or if you prefer not to go through the directory. MCP is the open standard AI assistants use to connect to tools like Paramark. Check your assistant's help pages for whether your plan allows custom servers. The server address is:

https://api.paramark.com/mcp

• ChatGPT: on the plugins page choose Add custom MCP server, paste the address, and pick OAuth as the sign-in method.

• Claude: under Connectors choose Add custom connector, name it Paramark, and paste the address. Leave the client ID and secret empty.

• Claude Code: run claude mcp add --transport http paramark https://api.paramark.com/mcp, then /mcp to sign in.

• Other tools such as Cursor or Codex: add a streamable HTTP server with that address and no API key. The tool opens the Paramark sign-in page the first time it connects.

Best practices & tips to get value from Paramark's MCP

What it is?

• A read-only way to get your MMM results, live MMM, experiment results and the data behind your models into ChatGPT, Claude and other tools that support MCP.

• A way to put your Paramark data to work in other places: building presentations, drafting messages, or combining it with other data and workflows.

• Scoped to your own account. The assistant sees what you see in the Paramark dashboard, and nothing more.

• A beta, available on the Advanced and Enterprise plans.

What it is not?

• A replacement for Leo. It isn't built to interpret your results, explain why they changed, or recommend what to do next. Ask Leo in Paramark for that.

• A way to change anything in Paramark. It cannot edit a model, a report, a plan or any other setting.

• A way to see data outside the Paramark account you signed in to.

• Generally available yet. It is in beta, so expect it to keep evolving.

Ways to use your Paramark data

Start with "Am I connected to Paramark?" to confirm which Paramark account the assistant is reading. Then tell the assistant what you want to pull or build, in plain language.

You can also ask follow-ups, such as "put that in a table" or "compare that with the previous report", and the assistant reuses what it already fetched. For questions about why results changed or what to do next, ask Leo in Paramark.

Tips

• Name the model if you have more than one: Accounts with several MMMs get asked which one you mean. Saying "the purchases model" or "the US model" up front saves a step.

• Say which period you care about: "Last month", "Q3" or "the last 12 weeks" all work. Without a period, the assistant uses the summary window of the latest report.

• Use Leo for interpretation: Use the connector to get your data into other tools, and Leo in Paramark to understand what it means.

• Check the numbers before you share them: AI assistants can make mistakes when reformatting or summarizing data, so compare anything important with your Paramark dashboard.

How can I try Paramark's MCP?

If you'd like to try Paramark's MCP, reach out to your Growth Advisor or post in your shared Paramark Slack channel. We can confirm whether MCP beta access is available on your plan and for your workspace and help you get started.

How can you provide feedback on Paramark's MCP?

Paramark's MCP is in Beta, which means your feedback genuinely shapes where it goes next. We're actively iterating on the data Leo can access, the questions it handles well, and the insights it surfaces, and we want to hear from you.

The fastest ways to reach us:

• Message your Growth Advisor directly, or drop a note in your shared Slack channel.

• Tell us what's working, what's confusing, and what's missing — especially the moments where Leo helped you move faster, or left you wanting more.

Thanks for helping us build it. This is just the first version, and the best ideas for where it goes next will come from the people using it every day.