The AI document platform Mintlify has completed its Series B funding round with a valuation of $500 million, with nearly half of its document traffic coming from AI agents.
According to Beating by Dongcha, AI document generation and maintenance platform Mintlify has completed a $45 million Series B funding round, valuing the company at $5 billion. The round was led by a16z and Salesforce Ventures, with participation from Bain Capital Ventures, Y Combinator, DST Global, and others, bringing the total funding to $67 million.
Mintlify uses AI to directly generate technical documentation from code, automatically syncing content with product updates. The platform currently serves over 20,000 companies, including customers such as Anthropic (for Claude Code documentation), PayPal, Coinbase, Microsoft, and Amazon. Founded by Cornell University alumni Han Wang (Co-founder and CEO) and Hahnbee Lee (Co-founder) in late 2022, the company underwent 8 product pivots before finding its direction. By early 2026, the company had reached a revenue in the tens of millions of dollars, with the main income coming from usage-based billing for advanced features, such as AI-powered chatbots embedded in customer websites.
One key metric shared by Wang illustrates why the company received a $5 billion valuation: Nearly 50% of all customer documentation views on Mintlify have come from AI agents rather than humans. These AI agents do not learn about the product from marketing pages but instead understand what the product can do, how to call it, and how to interact with it through the documentation. In Wang's words, "If you can't explain how to use the product properly, it's almost as if the product doesn't exist." Documentation, once the least favorite task of developers, is now becoming the infrastructure for whether a product can be discovered and utilized by AI. Mintlify's next step is to expand from public technical documentation to enterprise internal knowledge bases, following the same logic: Internal AI tools in enterprises also depend on structured knowledge sources. If the underlying knowledge is fragmented or outdated, the answers provided by the agents will be incorrect.
