Skip to main content

Connect External Systems

Everything that brings outside data into a conversation lives here: the connection to a provider, the endpoints built on top of it, and the saved sample responses you design and test against. This is the layer beneath your data points — when a flow needs a real order status or fires a real create-shipment call, this is where that data comes from.

How the layers fit together​

It helps to picture three layers, outermost first:

  1. Integration — the connection to a provider (Shopify, Freshdesk, your ERP): base URL, authentication, credentials. Defined once, reused everywhere.
  2. Custom API — a specific endpoint built on an integration: "fetch order by id", "create a ticket". Has its own input fields and response mapping.
  3. Data point — the named value or action a flow uses, usually backed by a custom API.

A conversation reads a data point, which calls a custom API, which runs over an integration. Ask the Copilot to work at whichever layer your request is about.

Integrations​

Connect a provider or change its credentials, auth, or base URL. Use a preset when the provider is built in (Shopify, Unicommerce, Flowcall, Shiprocket, Delhivery, Shadowfax, Openleaf, Xpressbees) — the Copilot already knows its shape. Use a custom integration for anything else by describing its authentication and a test call.

"Connect Shopify using these credentials." "Create a custom integration for our internal ERP with bearer-token auth." "Change the Freshdesk integration's base URL." "Rotate the API key on the Shiprocket integration."

An integration is defined once and shared by every endpoint built on it, so changing credentials here updates every custom API that uses them.

Custom APIs​

Create or update an endpoint, edit its fields or value formatting, or save provider API documentation for later. A custom API can draw from different sources — a preset integration, a custom integration, or even several integrations at once when a single call needs data from more than one provider.

A custom API can also be set as the source for built-in data — order details, order list, customer, product list, or order timeline — so that data comes from a provider other than your default ecommerce store. This is how you point Flowcall at orders that live in Unicommerce or a custom ERP instead of Shopify.

"Create a Shopify endpoint to fetch orders." "Use our Unicommerce order API as the source for order details instead of Shopify." "Set get_orders_erp as the order-list source." "Add a cf_reason field to the freshdesk_create_ticket API." "Strip the # from the order id before sending it to Shopify." "What are the Freshdesk ticket custom fields?"

What to include when creating an endpoint:

  • The provider and integration it should run on.
  • What it does — fetch, create, update — and against which resource.
  • Its inputs — what values get passed in (order id, pincode, ticket fields).
  • A sample response, if you have one — it lets the Copilot map fields correctly the first time.

The Copilot also answers provider-metadata questions (custom fields, available parameters) so you don't have to dig through the provider's docs yourself.

Sample data​

Save, fetch, preview, or explain saved samples in the Response Library. Saved responses matter for two reasons: the Copilot maps and reasons about your data more accurately when it can see a real payload, and your tests can mock API calls with a saved sample so they run without hitting the live provider.

"Save this order ID's response as delayed order." "Fetch order details for order number 1234." "Show me what get_order_status returns for a cancelled order."

Give samples clear names (delayed order, cancelled order, cod order) — those names are how you'll pick them later when mocking a test scenario.