Mintlify launches Base, a Git-based wiki that agents keep up to date

Why company wikis go stale

In a post announcing the product, Wang describes a familiar pattern. Someone writes a page, someone else reads it later, and in between nobody keeps it current. The people who know the most are the busiest, and checking old pages is no one's job. The product, the team and the plan change, but the page does not.

According to Wang, that is how "every wiki dies": one out-of-date page at a time, until nobody trusts it and everyone just asks in Slack.

He says agents make the problem worse. They only know what has been written down, and when an old page and a newer Slack thread disagree, an agent has to guess. A smarter agent guesses better, Wang writes, but it sounds just as convincing when it is wrong. He also cites Mintlify's own figures: on the public docs it hosts, more than three-quarters of traffic now comes from agents, and he expects internal knowledge to follow the same path.

Treating knowledge like code

Wang says that adding AI on top of existing tools, such as a question-answering sidebar, a summarize button or a connector for search, helps agents find pages but does not make the pages correct. In his words, finding the page was never the hard part, keeping it true is.

Mintlify's answer is to borrow the practices engineers use for code. The company keeps knowledge as files in Git, which gives every page a full history, lets every change be reviewed, and keeps the files in the customer's hands. According to Wang, Claude Code already edits files like these, and other agents can reach them through an MCP server. Because people and agents work from the same place, an agent that answers a support question can also update the page it used.

Wang also points to a controlled test run with a large engineering team using Claude Code. In that test, agents working from structured docs gave 64% more precise answers and used about half the tokens compared with raw code, according to Mintlify.

Agents as writing partners and maintainers

Wang wants the writing process to resemble how many engineers now code. The agent should sit beside the author as a copilot and thought partner, pulling in context from other pages, pushing back when something does not add up, and flagging contradictions with existing content.

He wants agents to handle upkeep after publication as well. He gives an example: a pricing change ships on a Friday, and by Monday an agent has drafted fixes for every page that mentions the old price and flagged a sales playbook that no longer matches. A person approves the changes before lunch. Wang says people will still decide what is true, but will not have to type it all up. He expects agents to take over most of this upkeep at companies over the next couple of years, and notes that Mintlify's Autopilot already does something similar for documentation as code ships.

What Mintlify Base is and where it stands

Mintlify built Base for its own use first. For the past three months, the company says, it has been using the tool internally, with nearly a thousand pages in one place where staff and agents work from the same files. Base is now available in beta, and Mintlify says it wants to shape the next version together with the teams using it. More information is at mintlify.com/base.

The pitch rests on a broader claim from Wang: the companies that move fastest will be those that get the most out of agents, which means putting them to work and giving them the right information to do it well. The source does not give pricing or general availability details for Base.