Late beta · Free Starter plan →Oprag is in late beta. The Starter plan is free — no credit card. →

Product

Discovery moves from browsing to asking.

Your visitors already know what they need. Oprag lets them say it, answers from what you've published, and shows you the intent your analytics never recorded.

The gap

You've already written the answer. They just can't find it.

Most companies don't have an information problem. They have a retrieval problem — and it's the visitor doing the retrieving.

What you've published

  • Marketing site
  • Product pages
  • Documentation
  • Help center
  • Developer portal
  • Knowledge base
  • PDFs and datasheets

What a visitor does

  1. Searches "pricing"
  2. Opens three pages
  3. Still doesn't know if it works with their stack
  4. Leaves

You'll see all four of these in your analytics. You won't see the reason.

The information was there the whole time. The visitor just had no way to ask for it.

What you learn

Analytics tell you where they went.
The conversation tells you what they were trying to do.

Three page views and a bounce look identical whether the visitor found what they needed or gave up. The sentences they typed are the only place the difference is recorded.

What your analytics record

/pricing        0:42
/docs           1:15
/integrations   0:19
exit

A session. Three pages. No conclusion available.

What the conversation recorded

  • what's the difference between the two paid plans?
  • is there a rate limit on the API?
  • do you work with Segment?

The same three pages, with the reason attached.

Signals, not certainty

Oprag shows you what visitors asked, in what order, and where the conversation stopped being useful. It doesn't tell you what anyone was thinking.

Patterns across conversations

One person asking about Segment is a question. Forty people asking is a gap in your product, your content, or both.

A layer, not a replacement

Keep your analytics. This sits next to them and answers the question they were never built to answer: why.

Getting live

There is no AI build to start here.

No model to choose, no pipeline to maintain, no infrastructure to budget for, and nobody on the team who now owns it.

  1. Add your content

    Paste your website URL or drop PDFs and FAQs into a project. Give a website source a re-crawl interval and it re-reads itself on that schedule, so when you publish, the answers follow.

  2. Set the look and test the answers

    Match brand colors, logo, and lead capture settings, then ask it questions in the dashboard until the answers are right — before a single visitor sees it.

  3. Go live

    Add one script tag to your site header, or call the REST API from your stack.

What you never do

  • Choose a model
  • Build a retrieval pipeline
  • Run a vector database
  • Tune prompts to keep answers usable
  • Re-crawl your site by hand
  • Staff it

The correct amount of ongoing work here is none, and that's the whole point.

Start free

Building headless? Read the API docs →

Recently added

What's new

Updated

  • Answers verified end to end in 60 languages, romanized input included
  • A dashboard inbox to read and review every conversation
  • Reports showing your most-used documents and the questions that went unanswered
  • Your logo and colors on the widget, on every plan

See what your visitors would ask. Connect a page, ask like a customer, and watch the answer cite its source. Free, no credit card.

Start free