Launch HN: Theneo (YC W22) – Generate Stripe-Style API Docs
I’ve been working with APIs for the past seven years, and the documentation process has been consistently the most painful. Nobody wants to have to do it, it’s filled with gruntwork and repetition, and takes a ton of time. Developers are not technical writers and documenting is never their priority, let alone making docs look nice. But for the reader—often the very same developers, when they’re consuming others’ work—it makes such a difference. How many times have you just looked at the API doc and decided “nope not going to integrate”? That was the case for me and my team many times. Developer communities often complain that even the most well-known companies’ public API docs are incomprehensible, and internal API docs are neglected even more.
The gruntwork and repetition are clues that there’s room for software to do a lot of this work, and particularly for ML models to help produce the content.
We want to allow anyone to create great docs without having a whole team of writers and designers, so we researched what makes Stripe, Square, and a few other API docs so great. First, it was definitely the content. Second, it was the user-friendly interface: layout, colors, scrolls, how easy it was to find information and give feedback. We talked to developers and technical writers to understand their pain points. Then we took all that research along with Maria’s experience as an ML Engineer at Microsoft, Deepmind, Google, and Facebook to come up with Theneo, a Technical Writer AI Assistant.
Theneo is like having a technical writer next to you. You upload your API collection (JSON, YAML, etc.). Our models automatically load and analyze API requests, methods, endpoints, request body, parameters, etc., then run some quality checks (e.g. for grammar errors) and give you content suggestions for the section names and descriptions.
API docs generated by Theneo are user-friendly, beautiful and interactive. we automatically generate all of your API requests in multiple languages and make it interactive for the reader. Here is an example: https://app.theneo.io/demo/theneo
For internal API docs, we created several integrations that make it easy to create and maintain API documentation: Postman Integration - you can pull all of your collections and use them in Theneo; Github Actions - you can define how and when you want API docs to be updated automatically (https://github.com/marketplace/actions/theneo-api-documentat... Visual Studio Code (coming soon) - by the end of this month you will be able to create and preview your API documentation right in your IDE. We do not use your private API docs for ML training.
You can see a quick overview of how our setup works here: https://www.loom.com/share/792fc6008a914d068ff8b34994fbc467
We are coming up with a lot of new integrations and features to make it easy for developers to generate and maintain internal or public-facing API docs. We would love to hear from you, what you are currently using, what you love or don’t love about those solutions, what stops you from havin...
97 comments
[ 2.1 ms ] story [ 173 ms ] threadAlso, on an unrelated side note, SEO seems very important here (for public facing API docs), so I'd put some emphasis on that as well. Good search mechanisms are also of great help in API docs. So I'd try to knock that out of the park as well.
Inclined to disagree on this one. I feel like creating API docs is uncommon largely because the tools are bad (and what tools do exist tend to be expensive). I think success for this kind of company might involve creating a market for documentation by lowering the barrier to entry.
Also, I'm not sure what the ML does here, is it like input open Api model -> output doc?. I'm sure you have to still add information manually.
Congrats on the launch!
We would love to hear your feedback and see how we can improve the published API doc and make it more beautiful
Regarding the ML, what we have is a technical writer AI assistant, that analyzes the API collections, gives suggestions, and the goal is that in just a few clicks to have the entire API doc ready.
The words are wrapping across lines in the link you gave. "asc is ascending and sorts from a to z." Looks like:
asc is ascending and sort
s from a to z.
> is it like input open Api model -> output doc?
Yes. Our ML pipeline analyzes the input, suggests content for the doc, and analyzes the quality of the content i.e. how intelligible is the content in your doc. It is up to you if you want to include the suggested content in your API doc. Feel free to schedule a demo call if you are more interested in details.
I really need a quick way to test whether this holds any water, and I don't want to provide you with my details to do it. If it doesn't work for my simple example, it won't work for my larger project - and I don't want to waste that time to begin with.
The best way to beat skepticism is to let people try the tech out and see for themselves that it actually works. There's too much "hotdog/not hotdog" tech in this space already.
Edit: I see there's a tiny "try free" button hiding at the bottom of the page. After clicking that, you're asked to provide your personal details. Then you're redirected to a page where you're suddenly asked to pay 20$ and sign-up with your credit card. This is pure nonsense. Enough with these dark patterns already.
Requiring our users to create accounts allows us to better serve our customers and to limit abusive actions should we need to. There are a whole set of security issues to consider. It's not even a "trial" if there's not even an account. We simply chose to prioritize the customers' needs who care about the product enough that they would not mind signing up for a free trial.
As devs ourselves, we would be happy to give back to community. Maybe by open sourcing some parts in the future, when we have more free time.
The friction is there because that's the cheapest/fastest way to go live.
It is a leaky funnel, but most early products have leaky funnels.
You might also consider working on softening your tone.
Instead of "you should" which might be appropriate in a code review, consider using the language "have you considered."
ETA: It would be helpful if you would have an example of the APIs where you highlight what was manually done by a person and what was automatically generated. The example docs (https://app.theneo.io/demo/theneo/theneo_api_reference) look like any API documentation system, it isn't clear what's your product's added value.
Edit 2: I just watched Ana's video from your post here, that's not on your homepage. It is 10x better than your homepage at explaining what you do :) I'm sold.
So happy that you watched the video, we need to improve our landing page for sure, we've been so concentrated on the product itself, we forgot about the landing page :D
Just a single data point, but perhaps reconsider how you describe it.
We develop an open source observability and automation platform for Kubernetes and our docs are at https://docs.robusta.dev/master/
Key questions for me:
* Can we easily migrate from Sphinx and RST?
* What do the docs look like?
* Can we make documenting non rest APIs more interactive? (E.g our yaml automations)
Regardless, looks cool. Thanks
thanks, means a lot!
What is the advantage of using your product to what is available for free except some CSS tweaks?
Redocly's basic plan starts at $69/month billed annually - $828 per year, professional $300/month billed annually, and enterprise not listed.
Readme's startup plan starts at $99 per month per project and enterprise is $2000 per month/project, etc.
Theneo's basic plan starts at $20 per month and enterprise plan is $100 per month. There are obviously many other differences in these plans so feel free to go check them out yourself and compare pricing.
I dunno, my API stays the same, but because we have 25 employees instead of 5 we pay a 5x premium?
Unfortunately whenever I import my Open-API spec to Theneo I get a success message, but it never becomes a usable collection.
Can you help me out?
API documentation has always been a hassle, no matter which tool or service I've tried they always tried to solve the small, isolated issue, instead of an end-to-end solution.
Theneo's promise to use ML to automate the most painful steps of a technical document is refreshing and exciting. I wish you the best of luck and hope this is a beginning of a new platform of a full-stack solution instead of an ad-hoc one.