Monjiro

The Monjiro concept

We believe the value of documentation lies not in how much information it contains, but in whether it can answer readers' questions


In practice, though, readers often can't find the information they need, or what is written has become outdated, and their questions are left unanswered.

We built Monjiro with two aims: to provide documentation that is easy for readers to understand, and to let the people who run the documentation easily keep updating it and keep publishing accurate information.

From now on, not only people but also AI agents will read documentation to do their work. For AI to do its job well, the standards, rules and other information it relies on as the source of truth must be carefully managed as documentation and kept up to date and free of contradictions.

To keep creating living documentation that both people and AI can rely on, Monjiro values the following four ideas.

01. A simple design that keeps the focus on the documentation

The documentation is always the main character. Instead of adding decoration or complex controls, we aim for screens that let readers focus on finding, reading and understanding information.

02. An AI chat that doesn't keep you waiting

Readers should be able to ask the moment a question comes up, and get an answer right away. Our performance target is for the answer to start appearing within 3 seconds of asking, and we put weight on how fast it feels.

With BYOK, the speed also depends on the response time of the AI service you use.

03. Designed to coexist with AI agents

Documentation is not read only by people. To make information easy for AI tools and AI agents to work with, Monjiro provides the following.

  • MCP integration: AI tools such as Claude Code can read your documentation and create pages. For details, see MCP integration.
  • llms.txt: The list of public pages is published in a format for AI. For details, see Publish a project.

Going forward, we are also considering an MCP endpoint that lets readers' AI tools read the public site, and support for new standards such as WebMCP.

04. A system for living documentation

Documentation is not finished once it is written. When operations change, the content needs to change too. In Monjiro, operators can review the questions readers asked the AI chat and how they rated the answers. We aim for a system where you can see where readers get stuck and what information is missing, and keep updating the documentation to match how things actually work. For details, see AI chat.