Replicate

AI API token

Run machine-learning models on Replicate, and manage predictions, deployments and trainings, from any Windmill script, flow or app.

Set up in 3 steps

Copy your API token

At replicate.com/account/api-tokens, copy your token.

Add the resource in Windmill

Open Resources, add a resource of type replicate, and paste the token.

Run your first script

Fork any script below and read before writing anything.

Scripts

Building blocks, if none of the projects above fit. Fork one, run it, or call it as a step inside your own flow.

Resource type

replicate — the shape of the credential every script on this page expects.

View raw schema →

Field	Type	Required	What it is
token	string	Yes	API Token retrieved from the dashboard.

Questions

Predictions are asynchronous

A prediction call returns immediately with a status. Poll it until it succeeds, or pass a webhook so a flow resumes when the model finishes.

Inference is billed by the second

Every prediction costs compute time. A flow that retries a slow model can spend more than you expect — cap the retries.

Is my token stored on the hub?

Never. It lives only in your own Windmill workspace, encrypted at rest.

Set up in 3 steps

Copy your API token

Add the resource in Windmill

Run your first script

Scripts

Update a deployment

List trainings

List public models

List predictions

List model versions

List files

List examples for a model

List deployments

List collections of models

List available hardware for models

Get the signing secret for the default webhook

Get the authenticated account

Get a training

Get a prediction

Get a model version

Get a model's README

Get a model

Get a file

Get a deployment

Get a collection of models

Download a file

Delete a model version

Delete a model

Delete a file

Delete a deployment

Create a training

Create a prediction using an official model

Create a prediction using a deployment

Create a prediction

Create a model

Resource type

Questions