🚀 Scaling pieces with AI — a rolling announcement

If you head to our Pieces page and sort pieces by Popularity, you’ll notice that Webhook and HTTP are used a lot.

So what?

HTTP and Webhooks are widely used. This means that although we have ~200 pieces, they’re not I’m not saying that we can actually get rid of HTTP and Webhooks (we’re also working on improving their experience every day), but we want to align our offering with our typical no-code users.

Announcing AI to Create Pieces

We’re positive that AI will be leading the creation and development of our pieces in the future, and today we’re starting with the first steps.

Here is how our initial roadmap looks like for this project:

  • Create piece authentication from OpenAPI specs.
  • Add a generic action to the piece: “Custom API Call”.
  • Test the authentication and the generic action pieces across tens of the pieces we built manually and compare results.
  • Make this available as a CLI tool for pieces developers (and keep shipping new features to this CLI).
  • Generate actions from paths in the OpenAPI specs.
  • Find ways to filter out meaningful actions.
  • Find ways to make the actions usable on the UI (not just 1:1 mapping of the API endpoints). For example: handling dynamic dropdowns, dynamic props and more.
  • Improve developer experience to ensure the tools reduce building times by X%.
  • Find ways to find meaningful triggers and build them (Scheduled and Instant).

Project Ownership

Our team member Saif owns this project and is doing all the work now, he’ll be updating this topic with progress and findings.


You can share your thoughts here or contact Saif or me to collaborate more on the project.


Our 1st Update - Scale Pieces :rocket:

You will be able to pick any openAPI spec. either from a URL or locally as a JSON file.

npm run cli pieces scale http://localhost:3000/slack.json

  • It will create a normal piece using ActivePieces CLI
  • Read your openAPI spec and prepare it for data extraction.
  • Extract the right fields for piece authentication.

Example Output:

export const SlackAuth = PieceAuth.OAuth2({
  description: '',
  authUrl: 'https://slack.com/oauth/v2/authorize',
  tokenUrl: 'https://slack.com/api/oauth.v2.access',
  required: true,
  scope: [

Using AI for extracting the needed data from openAPIs, You will be able to create your pieces easily

1 Like

Nice really love this idea!! And I can see this work.

What about dropping an API reference url from an application into AI, based on the different kind of API reference formats out there you could create a generic flow that will recognize the system and know where to look for. In this way you can just drop the API reference URL and AI will scrape out all the data it needs to set-up all actions and triggers, ask for or not asked for by the community :wink:

Might sounds to easy but having an ambition is always good haha

1 Like

Great idea Bram!
We were thinking about something similar to that. But for now, we’re trying to publish this version which depends on openAPI specs. as a first stage

Wait for it :pinched_fingers::rocket:

1 Like

:rocket: [BETA] Generate pieces using AI :rocket:

I’m here to announce that we’ve released the documentation for our CLI, which allows developers to generate pieces using AI and openAPI specifications.

Current Features

Using AI, you can now:

  • Automatically generate piece definitions based on the openAPI specification.
  • Auto-generate piece authentication from the openAPI security scheme.
  • Include a standard customApiCall action in every piece generated.
  • Automatically generate actions based on the openAPI spec paths and parameters.


Check the docs for more information! :rocket: