Documentation is where users go when they are already trying to succeed with your product, and where they quietly give up when they cannot find the one line they need. The answer is almost always on the site somewhere, in a guide two pages over, in an FAQ, in a section they did not think to open. A search box helps if they guess the right words. A chat widget that actually reads your docs helps even when they do not.
Your docs are the knowledge base, so use them
The reason generic chatbots are useless on a docs site is that they do not know your product. muro's AI does something different: you point it at your published docs and it answers from them, grounded on the exact pages you already wrote and maintain. There is nothing separate to keep in sync, when you update a doc you update what the AI knows. Paste your docs URL and it reads the pages and builds the knowledge base for you.
Answer in context, on the page they are on
The value of chat on docs is proximity. The reader is on the authentication page, stuck on token refresh, and instead of opening a new tab to search or digging through the sidebar, they ask in the widget and get the answer from your own docs without leaving the page. For a lot of how-do-I questions, that is the entire support interaction, resolved in one message, grounded in content you already published.
Drop it on any docs stack
One snippet covers the whole site. There are exact guides for the common docs stacks: Docusaurus, Ghost, Gatsby, and static sites built with Astro or SvelteKit; the full platform list is here. The widget is around 6 KB, loads async and lives in a Shadow DOM, so it will not fight your docs theme or slow the page.
Route real bugs to a human, or an agent
Docs traffic on a developer product produces two things: answerable questions and genuine bugs. The AI clears the first; the second needs a person, or in muro's case, optionally your own AI agent. Because there is a full REST API, webhooks and a hosted MCP server on every plan, your coding assistant can read a handed-off conversation, look at the bug and act on it in the same session. The MCP guide covers wiring an agent into the inbox; the API docs cover the rest.
Ship it
Add the snippet to your docs, point the AI at the published pages, and let email fallback catch anything for a human. Readers get answers from your own docs without leaving the page, the bugs reach a person, and the questions the AI cannot handle tell you exactly which docs to write next. Try it in the live demo, then add it to your site.