Polar Analytics MCP server icon

Polar Analytics

by Polar Analytics

HIPAA CompliantSOC2 ReadyISO 27001 Ready
Data & Research11 tools

Query unified Shopify, Meta, Google and Klaviyo commerce data from an AI agent through Polar's semantic layer. Anthropic lists 11 tools; Polar documents 8 of them with arguments. Read-only surface, OAuth or bearer API key, no application scopes.

Verified connector

Listed by Anthropic as a partner connector in its Connectors Directory.

Connection checked by Agentman on .

Anthropic states this reflects the level of review a connector received, not a security audit.

Connect Polar Analytics via MCP

https://api.polaranalytics.com/mcp

Works in any MCP-compatible client. In Agentman Studio it is one click — no config file to edit.

Polar Analytics Tools & Capabilities (11)

Tool names from Anthropic's directory listing. This server requires sign-in, so we could not read tool descriptions or parameter schemas.

Limits

  • No OAuth scopes exist. The metadata served at Polar's .well-known paths declares an issuer, authorization, token and registration endpoint and code_challenge_methods_supported, and omits scopes_supported entirely. There is nothing to diff a tool list against, and nothing at the grant separates one tool from another. The read-only property comes from the tool set alone.
  • The server answers .well-known/oauth-protected-resource with authorization-server metadata. Both that path and .well-known/oauth-authorization-server return a byte-identical document whose issuer is ai-service.polaranalytics.com. A bogus sibling path 404s cleanly, so this is not a catch-all — but it means Polar publishes no RFC 9728 resource descriptor, which is where a resource's own scope requirement would normally be stated.
  • Polar's enumeration covers 8 of 11 tools, and names one of the eight differently from the directory. get_dimension_values, get_metrics and get_custom_dimension_details are directory-only; we could not confirm them against a Polar source.
  • We could not read tool schemas or safety annotations. The endpoint returned 401 to an anonymous request, so parameter detail here comes from Polar's API article rather than the wire.
  • A customer-dimension breakdown returns personal data. Customer email, name, id, order id and postcode are all available dimensions, and no tool name warns of it. This is the sharpest edge on an otherwise read-only surface.
  • Polar's marketing overstates the connector. Its MCP page says the product can "trigger fixes automatically" and push actions into Slack. The connector does neither; an external automation platform with its own credentials does. Read that copy as describing a stack, not a tool.
  • Accuracy still needs supervision, and Polar says so. Its tips article instructs users to check every tool call, showing a case where Claude silently assumed a wrong date range, and recommends working inside a Project to reduce hallucinations. Verify the date range and metrics on the call before acting on the answer.
  • The connector is plan-gated and sync-gated. Polar Headless MCP is a priced line on Polar's pricing page, and a new workspace returns nothing for up to 24 hours during the first historical sync.
  • Whether a report deep link is shareable outside your workspace is unresolved. Polar documents generate_app_report_link as returning a link into the Polar app and we did not test it unauthenticated. Treat the sharing boundary as unverified rather than as closed.

Frequently asked questions

No tool among the 11 creates, edits or deletes anything in Polar. Dashboards, views and custom dimensions are all read-only through MCP — they are built in the Polar web app instead. Polar's own help centre states that Claude queries Polar in a read-only manner and that no trace of your chats is left inside Polar.

Yes, if you break a report down by a customer dimension. Polar's Shopify dimension list includes Customer email, Customer name, Customer id, Customer total spent and billing and shipping postcodes. Metrics are aggregates, but generate_report returns one row per dimension value, so a customer-level breakdown returns customer-level rows.

No. generate_report is a structured query builder, not a SQL passthrough. It takes typed arguments — metrics, dimensions, date range, ordering and filter rules — that are validated against Polar's semantic layer. Polar does sell direct SQL access to its Snowflake warehouse, but that is a separate add-on with no MCP tool.

Not through MCP. No tool schedules anything or sends a message. Polar's documented automation pattern puts the outbound step outside the connector: an n8n or Make schedule triggers a model, the model reads through Polar MCP, and a separate Slack or Gmail node with its own credentials sends the result.

None that narrow anything. Polar's authorization metadata declares an issuer, authorize, token and registration endpoint, and omits scopes_supported entirely. There is no scope naming a verb, so the consent screen offers no way to grant one tool and withhold another. The read-only boundary comes from the tool set, not from OAuth.

Anthropic's directory lists 11. Polar's own API article documents 8 with full argument lists, and one of those is named generate_report_link where the directory says generate_app_report_link. Four directory names appear in no Polar document, so treat 11 as the listing's count rather than a figure both parties confirm.

Effectively yes for ongoing use. Polar's pricing page lists Polar Headless MCP as a priced product line rather than a free tier, and its migration guide references a free 7-day trial that converts to a paid plan. Polar also states initial historical data sync can take up to 24 hours before the connector returns anything useful.

Whichever workspace you are currently logged into in Polar. Polar states the MCP loads data from your active workspace and that you switch it using the workspace switcher in the Polar app, not from the agent. Agencies holding several client tenants should confirm the active workspace before trusting a figure.

Sources

Use in Agentman

Connect once and your agents call these tools on their own — on a schedule, in a workflow, with nobody at the keyboard.

Open in Agentman Studio

Server Info

Category
Data & Research
Developer
Polar Analytics
Tools
11
Domain
api.polaranalytics.com

Using Claude Desktop or another MCP client? Setup docs — the connection URL above works anywhere.