“Verify that the session expires after 30 minutes of inactivity.”
Sounds right. But is 30 minutes your rule, or a number the AI has seen a thousand times? To check, you have to go and find the spec yourself — the work the tool was meant to save.
Canon
Test cases written from your specs. With the source attached.
Add your requirement docs, specs and standards to Canon. Hawzu's AI then writes test cases from what your documents say, and every case shows the document and section it came from. Checking one takes seconds instead of a trip to find the spec.
Reads PDFWord (.docx)Excel (.xlsx)CSVMarkdownPlain textLogs
Won't Scanned PDFsImagesVideo
“Verify that the session expires after 15 minutes of inactivity.”
Same sentence, plus a source tag. Hover it to read the text the AI used. Checking it is a ten-second look at a document you can open.
Add a document once. The AI uses it from then on.
Canon isn't somewhere you go to read documents. It's what the AI reads from. It's on by default, and you can switch it off for a single run.
Upload your documents
Upload a spec, or a whole folder of them — the folder structure comes with it. You can also drag files straight onto the page.
Hawzu reads it
It reads the text, then makes it searchable. Once the status says Grounding, the AI is using it. If a file can't be read, the page tells you why.
The AI shows its sources
Test case generation, Oracle, release readiness and Atlas all read from Canon, and each one shows which documents it used.
AI output you can check
You can always see where an answer came from, and you're told plainly when your documents don't cover something.
The AI follows your rules
Your PRD, API contract, accessibility standard or release exit criteria becomes what the AI writes against — not a general guess about how apps like yours usually work.
Every test case shows its source
Each generated case carries a tag for every document it used. Hover the tag to read the exact text the AI was given.
Check before you generate
Ask Oracle what your documents say about a topic. If it finds nothing, you've learned the spec is thin before you generate forty test cases, not after.
It tells you when a file can't be read
A document that failed to load is worse than a missing one, because the list still looks full. Every document shows its status, and a failed one shows the reason.
Four places your documents show up
Canon has one page of its own. Its value shows up everywhere else.
Generated test cases
The AI test case generator writes from your documents and tags each case with its sources. When it finds a gap — something the spec never says — the gap points at the document too.
AI test case generator →Answers from Oracle
Ask Oracle “what does our spec say about expired cards?” and it answers from your Canon documents, with the source listed underneath.
Oracle →Release readiness
The AI summary on a release judges it against the exit criteria you wrote down, and names them — so a reviewer can open the actual standard.
Releases & sign-off →The application map
Atlas can read your documents and suggest the areas and screens of your product. Each suggestion shows the document it came from, and nothing is added until you approve it.
Atlas — the product map →What Canon won't do
Better to know now than to find out halfway through a review.
It reads files you upload — not your wiki
PDF, Word, Excel, CSV, Markdown, text and log files. There's no Confluence connector and no website crawler.
Scanned PDFs won't work
A PDF made of page images has no text to read, so Canon turns it down and says why. Export a PDF with selectable text instead.
It doesn't watch your documents for changes
When a spec gets a new version, upload it and remove the old one, so the AI reads the current rules.
It can't test what nobody wrote down
If the spec says “users may edit shared steps” but never says what should happen, there is nothing to check. Canon won't invent the missing rule — the AI says the spec is silent.
“Where did this number come from?”
Someone in the review reads a generated test case that says data is kept for 30 days, and asks the question that stops every meeting. Without a source, someone has to go and look. With Canon, the case carries a tag — Data retention policy · §2 — so they open it, confirm, and move on. Then they ask about deletion, find nothing in Canon, and learn something more useful than a test case: nobody has written that rule down.
Canon is what it should do. Atlas is what it's made of.
Canon holds what your product promises. Atlas maps what it's made of — its areas and screens, what depends on what, and which parts no test reaches.
Add your spec. Check the sources.
Every feature on every plan. Free for five people, $20 per member after — no credit card.