Your reader has a question, not a keyword. The AI Assistant reads your guides, your API reference and your code steps, answers in their words, and shows the pages it used so they can check the working.
The writing half of this lives on the AI Agent →, and both are part of the AI features →
The answer is assembled from your pages and nothing else. No open-web guesswork, no half-remembered version of your API from two years ago, and no answer without the sources attached.
Guides, API reference endpoints, code steps and changelogs are searched together. Readers should not have to guess which shelf the answer is on.
Search runs behind SSO, JWT, magic links or a shared password. A reader gets results from the pages they can open, and no hint of the rest.
Requests go only to providers who do not retain your content or use it to train their models, and the sub-processors are listed in public.
A reader searching for “rate limit exceeded” does not know whether the answer is in the platform guide, the endpoint reference or the mobile SDK docs. They should not have to. Results are ranked by how well they answer the question, not by what changed most recently.
results appear as you type · no submit button
A code sample full of YOUR_API_KEY is a copy-paste bug waiting to happen. Signed-in readers see their own values, in the docs and in the search results.
Placeholders are replaced with the reader's real credentials and endpoints, so the first command they run is the right one.
Set a variable in your docs and every page, sample and search result renders it with the value that belongs to the reader looking at it.
The sidebar is for browsing. Everything else is a keystroke: readers and editors hit the quick switcher and jump straight to a page, an endpoint or a section.
Press ⌘K and jump to any page, endpoint or section.
No submit button, no page reload, no spinner to watch.
All localised versions searched together, best match wins.
Behind SSO, JWT, magic links or a shared password.
One bar across every project you run, on Enterprise.
Page tags drive filters and the related pages readers follow.
Terms defined once, with hover definitions inline for readers.
Generated llms.txt, plus an MCP server so AI clients can read your docs.
Search analytics show the top queries, the no-hit rate, the click positions and six months of history. The queries that return nothing are the most honest backlog you will ever get, and the agent will draft against it.
Keyword search ships on every paid plan. The AI Assistant and search analytics are on Pro and Enterprise, with a monthly grant of AI credits included. Searching across every project at once is Enterprise.
Gate the whole portal or a single section. Results are filtered per reader, so nobody learns a page exists by searching for it.
Wrap a block in a condition and it stays out of the index for everyone else. Preview as any audience before you publish.
Where there is a choice of model, you make it, and the hosting region of each one is shown at the point you choose it.
Fourteen days free, no card. Switch search on, watch the queries arrive, and close the gaps before they turn into tickets.
