Skip to main content
You decide exactly what a connected AI agent may do in Zippendo: which organization it works in and which permissions it has. The connector URL (https://api.zippendo.com/mcp) is protected by OAuth 2.1. You never paste a token: your client finds the sign-in from the URL, opens your browser, and you sign in to Zippendo and approve access. This page explains what happens and what you control.

​
The connection steps

When you add the URL to a client and connect for the first time:
1

Your client finds the sign-in

The server answers the first request without credentials with a pointer to its OAuth details. Your client registers itself automatically (no client ID to create) and opens your browser.
2

You sign in to Zippendo

If you’re not signed in already, you land on the Zippendo sign-in page. Sign in as you normally do.
3

You approve access on the consent screen

The Authorize access screen shows the client’s name, the Organization to connect and the Permissions it asks for. Choose the organization, tick the permissions you want to grant, and click Authorize. Click Deny to refuse.
4

You're connected

The browser hands control back to your client, which stores its tokens and refreshes them by itself. You don’t need to sign in again until you disconnect the agent or its access expires.
A team member who is limited to one brand can’t connect an agent. The consent screen refuses the request. Ask someone with access to the whole organization to connect it.

​
Choosing an organization

Each connection belongs to one organization. On the consent screen you pick which organization the agent may work in, and every tool call stays inside that organization. To work in another organization, connect again and choose the other organization on the consent screen. In clients that allow more than one server entry, you can also add a second connector with the same URL and approve it for the other organization.

​
Limiting permissions

On the consent screen you can grant fewer permissions than the client asks for, but never more. You must keep at least one. If the client asks for none in particular, the screen offers read and write access to everything. Grant only what the agent needs:
  • Only answering questions? Grant just the Read permissions.
  • Should it create and send shipments? Add write access to shipments (write:shipments).
The tools your agent sees are filtered to the permissions you approve. See Tools & scopes for the full map.
The agent acts as you, only in the organization and with the permissions you approved, and never with Zippendo staff rights. Every tool call runs the same sign-in, permission checks, plan limits and validation as a normal API request.

​
Tokens and refresh

Your client manages the OAuth tokens for you: You don’t handle these yourself: the client stores and refreshes them.

​
Disconnecting

To stop an agent’s access, go to Apps & integrations in the Zippendo app. The MCP (AI agents) card lists the agents you connected, with when each was last used and how many permissions it has. Click Disconnect next to the agent. Its access ends right away, including tokens it already holds. You can also remove the connector in your client (for example claude mcp remove zippendo, or delete it from the client’s MCP or Connectors settings). That stops the client from using it, but only Disconnect in Zippendo cancels the tokens on the Zippendo side.
The card shows only the agents you connected yourself. Each team member connects and disconnects their own agents.

​
For developers: discovery endpoints

The API is its own OAuth 2.1 authorization server. The standard discovery endpoints are public: A compliant MCP client handles all of this for you: you only ever give it the URL.