Read the docs from an AI tool
Fetch any doc or changelog page as raw markdown, or point an MCP client at the docs and changelog directly.
If you're wiring an AI assistant, IDE, or script to read this documentation instead of a human clicking through pages, you have two options: fetch raw markdown over HTTP, or connect an MCP client.
Fetch a page as markdown
Every product/API guide page and every changelog entry is also served as plain markdown. Append .md to its URL:
https://www.iadvize.ninja/docs/product/quickstart.md
https://www.iadvize.ninja/changelog/2026-06-30-docs-search.mdThe response is text/markdown: the page title and description, then its body. No HTML, no navigation chrome.
On the docs site itself, every eligible page has a copy button next to its title that copies this same markdown to your clipboard.
Generated API reference pages aren't included
Pages under /docs/api/** are generated from the OpenAPI spec and carry no
authored prose; there's nothing useful to export as markdown. Browse the API
reference directly for endpoint detail instead.
For a single file listing every page, see /llms.txt.
Connect an MCP client
The docs and changelog are also exposed as an MCP server, so a client like Claude Desktop or Cursor can query them as tools instead of you copy-pasting URLs.
The server is unauthenticated (it only serves the same public content as this site) and speaks Streamable HTTP at:
https://www.iadvize.ninja/api/docs/mcpAdd it to your client's MCP config:
{
"mcpServers": {
"iadvize-docs": {
"url": "https://www.iadvize.ninja/api/docs/mcp"
}
}
}It exposes four tools:
| Tool | Does what |
|---|---|
list_docs | List every doc page: title, url, description. |
get_doc | Fetch a doc page's markdown by its url. |
list_changelog | List changelog entries (date, title, url, type), newest first, optionally since a date. |
get_changelog_entry | Fetch one changelog entry's markdown by its url. |
list_docs and get_doc skip the same generated API reference pages the markdown export skips. Ask get_doc for one anyway and it returns an error rather than a near-empty result. For endpoint detail, browse the API reference directly.
list_changelog only returns entries meant for public visibility: a few entries are written for API integrators specifically and are left out of this list so it stays a clean, general-purpose feed. get_changelog_entry doesn't apply that filter, and neither do the entry's own HTML page or .md export: if you already have a changelog URL (from a link, a search engine, or anywhere else), fetching it directly can succeed even for an entry that never showed up in list_changelog. Don't assume every fetchable changelog URL was one list_changelog gave you.
The same publicly-listed set is also available as an RSS feed at /changelog/rss.xml, for anything that would rather subscribe than poll list_changelog.
Public content only
This server has no notion of your organization or account, so it can't answer questions about your agents, versions, or conversations. It only reads the same public docs and changelog everyone sees. For an MCP server that reads your actual conversations, labels, reports and agent versions, with nine read-only tools and three ready-made prompts, see Connect an MCP client to your dashboard, available to any organization.
Quality report
Answer quality from the thumbs-up/thumbs-down ratings on the assistant's replies (satisfaction, negatives by category, satisfaction by agent, and the most recent complaints), for a date range you choose.
Connect an MCP client to your dashboard
Point your own Claude or any MCP client at your AISA conversations, labels, reports and agent versions, authenticated as your organization. Nine read-only tools and three ready-made prompts.