Discover the AI Protocol – Tell Us What You Think!

Hi everyone,

I’m Thomas, the Engineering/Product Lead for the AI Protocol on Postman, and I’m excited to share that this feature is now live!

For those hearing about it for the first time, the AI Protocol is designed to make interacting with AI APIs as simple and intuitive as possible. It offers a streamlined interface where the majority of the API request setup is handled automatically after you select an AI model. This means you can focus on refining prompts and adjusting settings without the hassle of manually configuring HTTP requests.

This thread is your direct line to our team, and we’re eager to hear your thoughts, ideas, and questions. Your feedback is invaluable and will help shape the future of this protocol.

As we continue to enhance the AI Protocol, our immediate priorities include:

  • Refining error guidance to make troubleshooting easier and more effective.
  • Expanding model support to give you greater flexibility with AI tools.
  • Adding code export functionality to simplify integrating your requests into your workflow.

Let us know how you’re using the AI Protocol and what features or improvements you’d like to see. We’re here to listen and collaborate!

-Thomas

5 Likes

Hi @thomas-soria,

Congrats on the AI Protocol launch! Abstracting cross-platform LLM API access is valuable for different use cases.

I was trying out the AI Protocol (see Postman), and want to report a bug with variables (to securely handle credentials).

In terms of features, I’d like to see code generation and broader AI platform support, including Hugging Face open models.

Thanks,
Bernd

Thanks!

Regarding the bug, is the issue that your auth isn’t persisting when you re-open the item? We’re making some improvements to easy-auth and vault variable support in particular so those flows should improve in the near future.

And the two features you listed are indeed on our roadmap. Supporting arbitrary URLs for local or externally hosted models is a large endeavor and we’re working hard to get it right :slight_smile:

Hi Thomas,

Regarding the bug, if I use an API key (in a private workspace) it works fine. If I use vault {{vault:secret-name}} it works fine also. But if I use a variable {{secret-name}}, the value does not get replaced and I get a 401. This works fine in other collections.

Also, I recommend not to auto-save secrets (see screenshot, green message). One has to first enter a secret value, and then save it to the vault (this may take a few seconds). In a public workspace it is not secure. Persisting {{vault:secret-name}} after saving the secret value to the vault would be secure.

Thanks,
Bernd

1 Like

Awesome work with this;

We have to create a new collection when creating new AI Protocol requests; is there a way to add this to existing collections?

Thanks for all the awesome work!

Hey @ialimustufa :wave:

The AI request can been added to a Non-HTTP Collection, as we can do with the other Non-HTTP protocols.

These currently have certain limitations and less functionality as the HTTP Collection but we’re looking to address that in future releases.

1 Like