Mintlify’s $45M Series B Bet: Turning Documentation Into AI Infrastructure

In a funding climate where many software startups are tightening their belts, Mintlify just closed a $45 million Series B round led by Andreessen Horowitz and Salesforce Ventures at a $500 million valuation, pushing its total funding to $67 million. On the surface, it is another big-number announcement in the AI-adjacent gold rush. Look closer, and it is a wager on something more prosaic but far harder to replicate than model weights: the unglamorous documentation and internal knowledge that AI systems quietly depend on.

That distinction matters because regulators, enterprises, and workers are starting to discover the same uncomfortable truth: AI fails in the wild less because of raw model capability and more because it sits on top of fragmented, stale, and inconsistent information. Mintlify’s pitch is that this “knowledge layer” is no longer a back-office concern; it is becoming core infrastructure for AI agents inside companies and across the web. As lawmakers in Washington, Brussels, and Beijing debate how to govern powerful models, the real leverage may lie further down the stack, in who controls and maintains the structured knowledge these systems read from and act on.

From developer docs to AI system of record

Mintlify started in 2022 with a familiar pain point: developers stuck at 2 a.m. trying to integrate an API using a stale README and a GitHub issue from three years ago. The founders, Hahnbee and Han Wang, went through Y Combinator’s W22 batch and built a documentation platform that now powers docs for more than 20,000 companies and reaches over 100 million people annually.

For a while, the story was straightforward: make documentation beautiful, fast, and actually useful for humans skimming for code snippets and quick answers. Then the traffic shifted. Mintlify says nearly half of documentation traffic now comes from AI agents and AI-assisted workflows, not humans in a browser tab. That is a profound change: documentation is no longer just a developer convenience; it is becoming the machine-readable interface to how products work.

In this new environment, marketing sites are largely irrelevant to AI agents. They are designed to trigger human emotion and brand recognition, not to explain rate limits, edge cases, or API behavior in the precise ways models need. When an AI agent wants to know what a product does and how to use it, it goes to the docs. If those docs are incomplete or poorly structured, the product may effectively vanish from the AI ecosystem—an emerging discoverability problem that looks a lot like the early days of mobile and search, only this time the “user” is software.

Documentation as knowledge infrastructure, not content

Mintlify’s central claim is blunt: documentation is no longer just content; it is infrastructure. That framing aligns with what large enterprises are discovering as they bolt AI assistants onto sprawling internal systems. If an AI support agent is built on scattered help center articles, outdated pricing pages, and conflicting playbooks, it will confidently deliver wrong answers at scale. The model gets blamed; the real failure sits in the knowledge layer underneath.

The company has watched this play out in customer deployments. One example they cite: a company tweaks its pricing but neglects to update the help center. Every AI-based support agent trained on that content starts misinforming customers. This is not a hallucination problem; it is a data governance problem. As more organizations rely on AI to interact with customers, regulators are likely to start asking how systematically that underlying knowledge is maintained and audited.

Mintlify is positioning itself as the structured, continuously updated backbone that makes both external docs and internal knowledge usable by AI agents. The logic is simple: in an AI-native world, companies that invest in this layer will enjoy compounding advantages as machine-to-machine workflows become default. The rest will be stuck in a reactive posture, patching systems after bad answers hit users, regulators, or both.

Inside Mintlify’s AI-era product stack

The company’s product line reveals how this knowledge-layer thesis translates into specific tools. Mintlify’s original public-facing documentation platform remains the front door. Increasingly, however, customers are bringing the platform inside the organization to power internal knowledge bases, engineering handbooks, design systems, and best-practice repositories. The same patterns that make external developer docs effective—clear structure, versioning, and ownership—are being repurposed for internal AI agents that answer questions for employees.

To keep this content accurate, Mintlify offers “Workflows,” an always-on automation layer that helps keep documentation up to date across codebases and tools. It integrates with existing systems like project management platforms and source code repositories so that changes in one place trigger content updates elsewhere. It also provides its own AI agent to draft and update content based on prompts, an acknowledgment that even documentation about AI will increasingly be written by AI.

For external connectivity, Mintlify supports the Model Context Protocol (MCP), allowing customers to connect their knowledge directly to outside agents. In practice, that means organizations can expose structured documentation and internal knowledge to third-party AI systems in a controlled way, instead of hoping general-purpose models infer the right behavior from a marketing site. It is a small but important step toward a more interoperable AI stack, one where documentation behaves almost like an API for knowledge.