Skip to main content

MimicBot Docs

MimicBot is an embeddable AI chat widget that crawls your website, answers visitor questions with citations, and can complete contact forms and booking flows on behalf of your visitors. This site documents how to install, configure, and operate MimicBot in production.

Start here​

If you've never installed MimicBot before, begin with Getting Started. You'll install the widget, create your first bot, and verify your first crawl in under ten minutes.

What's in this documentation​

  • Getting Started — Install the widget with a single script tag, create a bot via the REST API, and watch the first crawl complete. Finish with a verification step so you know the widget is live before you send visitors to it.
  • Configuration — Theme the widget to match your brand, lock it down to the origins you control, tune the assistant persona that visitors see, shape the crawl so it covers the right pages, and add out-of-band sources for content that doesn't live on your public site.
  • Actions — Actions turn MimicBot from a question-answerer into something that actually does things on behalf of visitors. Learn the built-in action types, how to activate them on a bot, how to inspect execution history, and go deep on form replay — the flagship action that fills and submits HTML forms end-to-end.
  • Reference — The complete surface area: REST API, public widget endpoints, the widget embed API (script tag plus React <BotChat> props), authentication, rate limits, and the full error-code catalog.
  • Roadmap — Features that are partially built, feature-flagged, or planned — and what is explicitly not planned. Labeled honestly so you never try to use something that isn't production-ready yet.

How this site is organized​

The top navigation splits the docs into two sidebars: a Learn track (Getting Started, Configuration, Actions, Roadmap) that walks you through MimicBot in the order you'll actually use it, and a Reference track for when you need the exact shape of a request, response, or config field. Every reference page links back at the schema file it mirrors in packages/shared and apps/api, so you can always trace a documented field to the line of source it comes from.

Conventions used throughout​

  • Code blocks are copy-pasteable. Anything in angle brackets like <YOUR_TOKEN> is a placeholder you should replace.
  • API examples use curl. The same requests work from any HTTP client.
  • Fields flagged on the Roadmap are labeled inline so you don't accidentally build against something that isn't shipped yet.