Hey everyone!
A couple of weeks ago I was exploring how to integrate multiple LLMs into one project, and I had this idea: what if I could leverage AI to interact with APIs in natural language?
After chatting with my cofounder, we were both intrigued. So I opened up Cursor and hacked together a prototype.
After a few iterations, we had a working alpha: by dropping an OpenAPI file, we were able to get an AI Agent that understands it and can call those APIs using natural language.
The Agent understood every prompt, made the right API calls and even auto-corrected bad payloads.
For internal APIs, auth is always tricky, so we added a system that supports API keys and credentials client-side only. Tokens are never stored on our backend. We also integrated it with Postman, so you can easily import your collections.
We imagine different use-cases and plan to add integrations (Stripe, Slack, CRMs, etc) and allow users to embed it where they want (for example on their platforms, so that their end-users can chat with the APIs and perform actions).
Looking for early users and feedbacks.
0 comments
—
Show HN: I built a tool that turns OpenAPI specs into an agent · HackerTrans
After chatting with my cofounder, we were both intrigued. So I opened up Cursor and hacked together a prototype. After a few iterations, we had a working alpha: by dropping an OpenAPI file, we were able to get an AI Agent that understands it and can call those APIs using natural language. The Agent understood every prompt, made the right API calls and even auto-corrected bad payloads.
For internal APIs, auth is always tricky, so we added a system that supports API keys and credentials client-side only. Tokens are never stored on our backend. We also integrated it with Postman, so you can easily import your collections.
We imagine different use-cases and plan to add integrations (Stripe, Slack, CRMs, etc) and allow users to embed it where they want (for example on their platforms, so that their end-users can chat with the APIs and perform actions).
Looking for early users and feedbacks.