Using Downsizing with Make

Prerequisites

A Make account (make.com, any plan) with permission to edit scenarios, and a Downsizing account that can make requests (pay-per-use or a connected subscription both work). Grab your Inference ID and API key from the Inference API page.

Why not the native OpenAI / Anthropic Claude apps

Both the OpenAI app and the Anthropic Claude app's connections take only an API key — there is no field for a custom base URL, Azure-style endpoint, or proxy. Make's "Make an API Call" module on the OpenAI app is also fixed to a path relative to https://api.openai.com, so it can't be redirected either. Since neither app can be pointed at a proxy, use Make's generic HTTP app, which builds a raw, fully-configurable request and is the standard way to call any REST API from a scenario.

Manual Setup

1. In your scenario, add an HTTP module and choose Make a request.
2. Set URL to your Downsizing Anthropic-compatible endpoint:

https://api.downsizing.dev/YOUR_INFERENCE_ID/anthropic/v1/messages

3. Set Method to POST, then add these Headers:

Authorization: Bearer dws_YOUR_API_KEY
Content-Type: application/json

4. Set Body type to Raw / JSON (application/json), and use a body along these lines, mapping in fields from earlier modules where you see {{ }}:

{
  "model": "claude-sonnet-5",
  "max_tokens": 1024,
  "messages": [
    { "role": "user", "content": "{{Insert your prompt or a mapped field here}}" }
  ]
}

Replace YOUR_INFERENCE_ID with your workspace UUID from the dashboard, and dws_YOUR_API_KEY with your Downsizing API key. Set modelto whichever Claude model you want Downsizing to route — Downsizing selects the actual upstream model itself, so any recognizable name works. If you'd rather target OpenAI's wire format instead, POST to /YOUR_INFERENCE_ID/openai/v1/responses with an OpenAI Responses-shaped body — see the Inference API docs for the exact payload.

The HTTP module automatically parses the JSON response, so downstream modules can map fields straight out of it (for the Anthropic endpoint, the reply text is at content[] → items where type is text).

Verify

Run the scenario once (or use "Run this module only" on the HTTP module) and confirm you get a 200 response with a completion back. Then check your savings dashboard to see the request appear in your usage metrics.