llms.txt and browser agents
More of your readers are AI tools. Octri gives them a clean way to read your docs without any setup, and an optional way for agents to use them from inside a browser.
llms.txt
Your site publishes /llms.txt, a plain-text index of your pages for AI tools to read. It builds itself from your published navigation, so it stays current.
| SEO setting | /llms.txt |
|---|---|
| SEO on and indexable | Served |
| SEO on, not indexable | Served |
| SEO turned off | Not served (404) |
Turning off indexing keeps it available on purpose, because a site can hide from search engines and still want tools to read it. It's unavailable behind access control, along with the rest of the site. See SEO.
Copying a page for an LLM
Endpoint pages have actions at the top to copy the page as Markdown for an LLM, or to view the raw Markdown. A reader can paste an accurate description of your API straight into their assistant.
Browser agents
WebMCP lets an AI agent running in the reader's browser use tools your page offers, instead of working out the page by looking at it. It is off by default.
Turn on Offer WebMCP tools in Docs settings, under Browser agents. Every docs page then registers these tools with browsers that support WebMCP:
| Tool | What it does |
|---|---|
search_docs | Searches your docs |
get_site_overview | Describes the site and how it's organized |
read_page | Reads the current page or another page on the site |
navigate_page | Moves the reader's browser to a page |
The setting applies when you save. In a browser without WebMCP, nothing changes for the reader.
WebMCP helps an agent already in a browser tab on your docs. An MCP server connects an assistant like Claude or Cursor directly, and can call your API as well as read your docs. They do different jobs, and you can use both.