A lightweight AI-powered tool for generating OpenAPI specs. No login, no setup. Built by Kong to make spec creation and collaboration faster with AI. It’s a first step, with more planned around how AI can improve the full API lifecycle.
At Kong, we’ve been playing around with ways to make API design less painful. There’s been a lot of progress in AI dev tools lately: full-stack vibe coding tools, code assistants, you name it. These are genuinely impressive, but they’re designed to cover a wide range of use cases. At Kong, our focus is different. We build API infra and tooling like Kong Gateway, Konnect, and Insomnia, so when we think about generative AI, we're thinking specifically about how it fits into real API workflows and infrastructure.
create-api.dev is a simple experiment in this direction. You describe an API, get an OpenAPI spec back, and can start tweaking or sharing it right away. No accounts. No setup. It runs on our open source spec renderer and goes through Kong’s AI Gateway.
It’s early. Let us know what you build, what’s missing, or what direction you'd take it next.
Try it here: https://create-api.dev
Impressive initiative by Kong! 👏 The integration with Kong’s AI Gateway sounds promising. How does this tool differentiate in terms of performance compared to other AI-driven API tools? Any plans for additional features in the near future? Keep up the great work!
@jaredl Thanks for the compliment! RE Performance - lots of dimensions to consider, but the main ones we are optimizing for are
1) API spec quality - how do we know if the generated specs are "good" (and what does that even mean)? This is where we lean on eval frameworks like https://github.com/confident-ai/deepeval . As we iterate on the system prompt the eval tools help us make sure quality is moving in the right direction quantitatively (not just "seem better"). We also luckily have a small army of API design experts at Kong who are giving feedback and help us improve. With that said - let us know if you make a spec and have feedback.
2) AI Guardrails / Safety - very large topic, but the team built a multi-layered approach - leaning heavily on some out of the box Kong AI Gateway features
The team wrote in a bit more detail in our eng blog (including sharing our initial system prompt) and there will definitely be follow ups on these topics so stay tuned!
create-api.dev by Kong
Tidyread
Impressive initiative by Kong! 👏 The integration with Kong’s AI Gateway sounds promising. How does this tool differentiate in terms of performance compared to other AI-driven API tools? Any plans for additional features in the near future? Keep up the great work!
create-api.dev by Kong
@jaredl Thanks for the compliment! RE Performance - lots of dimensions to consider, but the main ones we are optimizing for are
1) API spec quality - how do we know if the generated specs are "good" (and what does that even mean)? This is where we lean on eval frameworks like https://github.com/confident-ai/deepeval . As we iterate on the system prompt the eval tools help us make sure quality is moving in the right direction quantitatively (not just "seem better"). We also luckily have a small army of API design experts at Kong who are giving feedback and help us improve. With that said - let us know if you make a spec and have feedback.
2) AI Guardrails / Safety - very large topic, but the team built a multi-layered approach - leaning heavily on some out of the box Kong AI Gateway features
The team wrote in a bit more detail in our eng blog (including sharing our initial system prompt) and there will definitely be follow ups on these topics so stay tuned!
BestPage.ai
Congrats on the launch! Instantly generating and sharing OpenAPI specs with AI is a real time-saver.
create-api.dev by Kong
@joey_zhu1 Thanks! let us know what you build (you can use the share link). and we'd love any feedback