The first person to use a new API is rarely the person who built it. That stranger arrives with a job to do, a deadline, and perhaps an error message that says little more than “unauthorized.” At Adobe’s Frame.io, the trip from reading the instructions to making a successful API call could take anywhere from one to fourteen days, according to a case study published by Fern. The API itself was not the whole problem. The team had no SDKs, examples were maintained by hand, and documentation had drifted away from the specification.
After Frame.io adopted Fern’s documentation and generated client libraries, the company says that first successful call took less than an hour. That is a customer-reported result, not a promise that every API will improve by the same amount. It does, however, reveal Fern’s premise: the welcome mat is part of the product.
- Fern generates API documentation and SDKs from a shared definition, then helps publish and maintain them.
- Its customers include Square, Auth0, Adobe’s Frame.io, Twilio and ElevenLabs; Fern reported more than 200 customers when Postman bought it.
- Postman acquired Fern in January 2026. The brand and product continued, and Fern later launched an agent for documentation work.
A specification is not a greeting
An OpenAPI file can describe endpoints with admirable precision. It does not, by itself, decide how a Python developer should page through results, how a Java client should handle errors, or whether an example in the docs still matches the library in the package registry. Those are small decisions until someone must make them in nine languages, on every release. Then they become an engineering department.
Danny Sheridan and Deep Singhvi knew what that department looked like. They met at the University of Michigan in 2015 and later encountered internal code-generation systems at Amazon and Palantir. Those companies could afford people to maintain client libraries and bindings. Most API businesses cannot. Fern, which they founded in 2022 and launched in Y Combinator’s Winter 2023 batch, was their attempt to bottle the useful part of that machinery.
Fern even had to fix its own front door. In a candid account of a 2025 marketing-site rebuild, its team said the old site failed to show how polished its output could be. The new one cost roughly $100,000 across design, animation, development and brand work. The company reported almost twice as many page views after launch. A later redesign shifted the site again, from explaining the generation pipeline to letting visitors see the docs and SDK experience. Selling the welcome mat required showing one.
The mechanism is straightforward enough to draw on a napkin. Give Fern an API definition. Fern builds a reference site and client libraries, then threads examples from those libraries back into the docs. The value is less in making any one page handsome than in keeping the pages, methods and published packages in agreement after the next change.
Fern Docs adds guides, an interactive API explorer, version switching, branded design and an editor for colleagues who do not want to open Git. The SDK generator produces language-specific clients with the unglamorous features that users notice immediately: retries, pagination, authentication and useful errors. Ask Fern answers questions inside the docs. These pieces can be bought separately in the market, but Fern’s argument is that the seams between them are where confusion starts.
The expensive part is the second version
Launching a single SDK is a project. Maintaining it while an API changes is a habit. Auth0 says it replaced hand-written client libraries with Fern-generated SDKs in seven languages and saved more than 7,000 engineering hours a year. ElevenLabs’s experience has a different tempo: its API changes multiple times a day, while its JavaScript and Python SDKs ship weekly and see more than one million combined downloads per week. The company says it chose predictable, deterministic generation, with custom code layered on top, rather than a pipeline that improvises each release.
The first thing to fail, in these accounts, was not raw server capacity. It was the connective tissue: Frame.io’s hand-maintained examples and absent SDKs; Deepgram’s old docs platform, which could not represent its WebSocket APIs through AsyncAPI; Auth0’s growing burden of maintaining libraries by hand. Each customer changed its mind when separate tools or internal maintenance became more expensive than a shared pipeline. The figures come from Fern’s customer case studies and describe those customers’ own circumstances.
“It’s one of those things that I’m very happy to buy, not build.”Paul Asjes, developer experience engineer at ElevenLabs, on SDK generation
The customer is the next developer
Fern sells to the company publishing an API, but its daily audience is the developer trying to integrate with that API. It also increasingly serves a reader with no eyes at all: a coding agent. Fern now serves machine-readable documentation, an llms.txt index and an MCP server alongside pages for humans. Ask Fern provides answers inside the site. In August 2026, the company launched Fern Agent, which can read search queries, feedback and analytics, draft changes, and hand them back for human review.
There is a useful irony here. The company that automated SDK writing still insists that its generator be deterministic. AI belongs where it can help diagnose a weak page or propose an edit; the released client library needs to be repeatable. In Fern’s account of building its agent, its own team reviewed the generated work before it landed. That is a less magical story than “press a button,” and a more credible one.
The business model follows the work. Fern publishes a free Docs tier, while enterprise customers pay custom prices for governance, scale, support, advanced design control and hosting options. SDK generation has self-serve and managed paths. Buyers can instead combine a docs platform such as Mintlify or ReadMe with an SDK generator such as OpenAPI Generator, Speakeasy or Stainless. A company with a strong in-house release team may reasonably choose that route. Fern’s pitch is that one definition and one workflow spare the API team from reconciling several outputs after every change.
Postman bought the front door
Fern raised a $9 million Series A led by Bessemer Venture Partners in April 2025, bringing announced total funding to $13 million. In January 2026, Postman bought the company. Neither side published the purchase price. Postman’s familiar territory is API design, testing and collaboration; Fern extends that work to the moment an outside developer sees documentation and installs a client library. Fern says its product and brand continue independently within Postman, and its team stayed to build them.
For readers building an API, the copyable idea is modest and demanding: treat the first successful call as a product metric. Keep one trustworthy definition; make documentation examples use the SDK people actually install; run generation in the release process; test the first integration from a stranger’s chair. Fern can automate much of that work. It cannot repair an unclear API contract, decide how a customer should be authorized, or make an incomplete specification complete. Someone still has to care about the welcome.
Customer figures are reported by the named companies in Fern’s published case studies and may reflect their particular API, release cadence and migration.