Files
PDFs, Word documents, slides, spreadsheets, CSVs, Markdown, plain text, images, MP3 audio, and video, up to 15 MB per file.
Sources
Get the information in without losing where it came from.
Upload files, import a web page, paste a note, or pull a video transcript. Everything arrives with its origin intact and stays inspectable afterwards.
Atlas Launch · Add knowledge4 ways in
Upload
Files up to 15 MB each
Website
Full-text web extraction
Text
Paste a note or excerpt
YouTube
Preview, then import a transcript
Website URL
Website extracted12 chunks · canonical URL retained
Recently added24 sources
Customer interviews.pdfPDFIndexed
Reliability review · Q3Web pageIndexed
Founder call transcriptYouTubeIndexed
Pricing teardown.pptxDeckQueued
Import activity
Website import
Extracted 4 pages, kept canonical URLs
2 minutes ago
Upload
Customer interviews.pdf · 38 chunks embedded
1 hour ago
Guided product demoReady
What you can do
Four ways in, one consistent record.
Web pages
Import a URL and Prism extracts readable full text, keeping the canonical link so citations resolve later.
Notes and excerpts
Paste text directly when the thing worth keeping is a paragraph, not a document.
Video transcripts
Add a YouTube URL in one step and Prism saves it as a playable, searchable transcript source.
Bulk URL import
Paste a list of links and bring in a whole reading queue in one pass instead of one at a time.
Typed on arrival
Filter by file type or your own tags to find the material you need.
Sources beyond your plan's semantic index cap still import successfully. Their metadata and text stay searchable while semantic indexing is paused.
Whole queues, not single links
Bring in a reading list in one pass.
Paste a block of URLs and Prism extracts each one, reports what succeeded, and files everything into the Project you chose. Failures are surfaced rather than silently dropped.
- One URL per line, no separate CSV or import template.
- Extraction handles pages that need rendering, not just static HTML.
- Every imported page keeps its canonical URL.
Import from a URL list
Paste one URL per line
https://example.com/engineering/reliability-2026
https://example.com/blog/launch-retrospective
https://example.com/docs/service-levels
3
queued
3
extracted
0
failed
Provenance kept
Every source stays traceable.
Prism stores extracted text next to the original reference. When an AI client cites a passage months later, the path from the claim back to the page it came from is still intact.
- Original URLs, file names, and import dates are retained.
- Prism prepares the text for search and keeps the original source.
- Indexing state is visible per source in Knowledge Health.
Source · Reliability review · Q3
Provenance
Original URLRetained
Imported12 Mar 2026
Source typeWeb page
Semantic indexCurrent
Extracted text12 chunks
Full text is stored alongside the original reference, so a citation can always be traced back to the passage it came from and to the page that passage was taken from.
Prism
Own your knowledge. Choose your AI.
Start a Project and add a file, link, or note. Ask about it in Prism, then connect another AI app whenever you want.