The Polar MCP (Model Connection Protocol) allows you to connect your AI assistant (like ChatGPT or Claude) directly to your Polar workspace so you can query your ecommerce and marketing data conversationally.
If your organization manages multiple workspaces — for example, different brands, stores, or regional accounts — the MCP can be used across all of them. However, it’s important to understand how to toggle between workspaces correctly to ensure the MCP is always referencing the right dataset.
In this guide, you’ll learn:
How the Polar MCP interacts with multiple workspaces
How to switch the MCP between workspaces
Best practices for keeping your data queries accurate and organized
How Polar MCP Works with Workspaces
The MCP connects to whichever workspace you’re currently logged into in Polar.
Each workspace in Polar operates independently, with its own connectors, metrics, and data models. When you activate the MCP, it automatically pulls data from your active workspace — the one you see in your Polar dashboard.
For example:
If you’re logged into the “US Store” workspace, MCP queries will return data for that store only.
If you switch to “EU Store,” MCP will refresh and source data from that workspace instead.
This flexibility allows agencies and multi-brand teams to manage multiple datasets seamlessly using a single MCP connection — without needing to reinstall or duplicate configurations.
Key takeaway:
The MCP always mirrors the workspace you’re actively using in Polar. Switching workspaces in the app automatically changes the MCP’s data source.
Switching Between Workspaces
Follow these steps to change which workspace the MCP is connected to:
Open your Polar account and ensure you’re logged in.
Navigate to the workspace menu in the top-right corner of your Polar interface.
Select the workspace you want to switch to (e.g., “Brand A,” “Brand B,” “Agency Demo”).
Once the new workspace loads, the MCP will automatically update to use this workspace’s data.
No additional setup is required.
You can confirm the change by checking your Polar workspace name in your MCP tool (ChatGPT, Claude, etc.).
Important note:
If you have both client and internal workspaces, always double-check which one you’re connected to before running a query. This ensures you’re not pulling sensitive or unrelated data.
Best Practices for Managing Multiple MCP Workspaces
Here are a few tips to keep your setup organized and your queries consistent when using the MCP with multiple workspaces:
1. Name your workspaces clearly
Use descriptive names in Polar like “Brand A – US” or “Brand B – UK” so it’s easy to see which data you’re viewing. This minimizes confusion when switching within your MCP tool.
2. Keep connector access up to date
Each workspace maintains its own integrations (e.g., Shopify, Meta Ads, Klaviyo). Make sure all relevant connectors are synced in each workspace before running MCP queries, or you may see incomplete data.
3. Log out before switching users
If multiple team members use the same MCP integration, always log out of Polar before switching to another workspace. This prevents cross-account caching and ensures the correct data source is loaded.
4. Refresh data regularly
If you recently added a new connector or updated metrics in Polar, reloading the workspace ensures MCP uses the latest data.
5. Verify results in Polar
When in doubt, validate any data pulled through the MCP against your Polar dashboard. This helps confirm you’re querying the correct workspace and data set.
The Polar MCP is designed to work seamlessly across all your workspaces — letting you switch between brands, clients, or regions with just a few clicks.
By understanding that the MCP always references your active workspace, you can ensure accurate, efficient data analysis across every account you manage.
Key reminders:
The MCP connects to your currently active workspace automatically.
Switching workspaces in Polar updates your MCP data source instantly.
Use clear workspace names and double-check before querying to avoid data mix-ups.
For further setup guidance or troubleshooting, see: