Skip to main content
Stripe tells Rippit what a customer pays and what happened to their money: the plan, the failed invoice, the refund, the dispute. Joined to conversation analysis, “these 40 tickets are about billing” becomes “these 40 tickets are about billing, and 15 of them come from customers with a failed payment this month.”

What the agent gets

  • Reads across the Stripe API with stripe_api_read: customers, subscriptions, invoices, charges, payment intents, refunds, disputes, balance transactions, and payouts.
  • Account details for the connected Stripe account.
  • Stripe documentation search, so the agent can look up how a Stripe object or field works.
  • Writes, only if you enable them: stripe_api_write can create, update, cancel, refund, and delete through any Stripe API method, and the Treasury tools can move money.

Setup

Stripe connections are available only to workspace admins. In Rippit, open Connections → MCP Servers, choose Stripe, and pick one of these identity options. There is no Stripe app to configure first.

Your Stripe account

Choose OAuth to sign in to Stripe and approve Rippit. On Stripe’s consent screen, choose the live accounts or sandboxes to connect and the permissions for each. The connection acts as the Stripe user who authorized it. If sign-in fails, check that a Stripe admin has not turned off MCP access under Settings → MCP and CLI access in the Stripe Dashboard. Live mode and sandboxes are set separately.

Stripe agent API key

Create a restricted key in the Stripe Dashboard under Developers → API keys → Create restricted key, and choose Authorizing agent access to your account so Stripe tags it as an agent key. Give each resource only the access the agent needs, usually Read. Stripe’s MCP server does not accept full secret keys (sk_) or restricted keys without the agent tag.

Identity and scope

Every request runs as the Stripe user who authorized the connection, or with the permissions of the API key. A resource the user or key cannot read is a resource the agent cannot read. Connecting a test sandbox first is a safe way to try the agent before connecting live data.

Write access and approvals

The connection starts read-only: write tools stay off until a workspace admin enables them under Tool permissions. Because stripe_api_write covers every Stripe write method, enabling it allows refunds, subscription cancellations, and invoice voids, not just one kind of change. Stripe adds its own check on top. Refunds and money movement need a person to approve them in the Stripe Dashboard, at a link the agent provides, within 24 hours. Use a restricted key with read-only permissions if the connection should never write, even if someone later changes the Rippit tool selection.

Tools

You choose which tools the agent can use under Tool permissions on the server’s page, one at a time or a whole group at once. See Setting one up. Which tools start enabled depends on how Stripe marks each one, so check Tool permissions after connecting and turn off anything you don’t want the agent to use.