Baseten

Baseten scaled its scrappy startup-level docs to agent-ready knowledge infrastructure without ever slowing down the product launches that made it one of AI's fastest-growing companies.

Success story snapshot

Challenge: Baseten is a leading AI inference company, the fastest moving category in software. They needed docs that could accommodate a full site overhaul, an ever-expanding API surface, and a new class of readers: coding agents.

Solution: Baseten has built on Mintlify since the company's early days, pairing a Git-based docs-as-code workflow with preview deployments, an interactive API playground, and Mintlify's AI-native layer, so both developers and their agents get accurate answers from a single source of truth.

Key Results:

• Full documentation redesign alongside a major product launch
• High throughput of pull requests merged through the docs-as-code pipeline
• Mintlify-generated MCP server shipped in Baseten's official agent toolkit

Documentation at the front door of inference

Baseten is the training and inference platform behind some of the most demanding AI workloads in production. They turn open-source, fine-tuned, and custom models into scalable API endpoints. Having closed two rounds that raised $450M in just four short months, the company serves a rapidly growing audience of inference engineers who evaluate infrastructure the same way they build it. They read the docs.

“For a company whose buyers are engineers, docs.baseten.co is the front door to our product,” says Philip Kiely, Head of AI Education at Baseten. "We've been using Mintlify for a couple of years at this point. I was a very early customer.”

Pushing the platform to its limits

A Mintlify customer since 2024, the team's day-to-day workflow is pure docs-as-code. Writers and engineers open pull requests against the docs repo, previews build automatically, and merges to main deploy to production in minutes.

An expanding API surface gets the same treatment. Baseten's inference API reference runs on Mintlify's interactive playground, so developers send real chat-completion requests without leaving the page. OpenAPI-driven reference pages, reusable snippets, and multi-language code tabs keep hundreds of fast-changing endpoints accurate from a single source.

In spring 2025, the team executed a complete overhaul of its documentation, migrating to a new Mintlify theme and rebuilding the site's structure in the weeks leading up to a major release. Preview deployments let the team pressure test every commit at a shareable URL before anything hit production, and the new docs went live alongside the launch. "We're definitely pushing the platform to its limits in terms of customizability," says Kiely.

Docs for agents

Baseten's users are increasingly sending their agents to retreive documentation and build based on it. Baseten has leaned into Mintlify's AI-native layer so coding agents can pull accurate documentation context on demand without guessing at URLs to web fetch. The company's official agent toolkit, baseten-skills, instructs every user to install the Mintlify-generated MCP server at docs.baseten.co/mcp right alongside Baseten's backend MCP.

Every page is served agent-ready through llms.txt and Markdown endpoints, and the team actively works with Mintlify to make the machine-readable surface as complete as the rendered one. On the human side, semantic search and the embedded AI assistant answer questions on the spot. Query analytics reveal exactly what developers are searching for. Questions like "how to run vLLM" or "how to run custom Docker images" show the team where to invest in content that gives their users the knowledge they need.

The bottom line

Partnering with Mintlify allowed Baseten to turn its documentation into launch infrastructure. A system that absorbs redesigns, ships alongside the product, and serves an audience that's half human, half agent. The two companies work together in a shared Slack channel where bugs get squashed and product feedback flows straight to the people building the platform.

When Baseten announced its Series D, the team credited its growth to customers, surging inference demand, and "incredible partners like Mintlify." It's how a company defining the inference era keeps its knowledge moving as fast as its models.