Give workflows your documents, not the internet.
Upload files or crawl a website into a knowledge base. A workflow step retrieves the passages that matter and the model answers from them, with the source attached.
Built in. Nothing to configure.
Files and websites
Upload PDFs, text, and office documents, or point a crawler at a public site. Every document is parsed and split before it is indexed.
Three ways to split
Fixed size, recursive, or sentence-aware chunking, from a hundred to eight thousand characters. Pick per base; the default is fine for most teams.
Meaning, keywords, or both
Vector search, keyword search, or a hybrid that merges the two. Return chunks or whole documents, with an optional size budget.
Check it before you trust it
Ask a question, see which passages come back and why, and tune the base until the answers are the ones a person would pick.
Per-base permissions
Each base has its own access level: none, use, manage, or owner. A workflow can only read a base its author can.
Every change is logged
Adding, re-indexing, or deleting a document writes an audit event, the same as every other change in the workspace.
Three steps, then it runs.
- Create a base and add sourcesDrop in files or a website address. Indexing runs in the background and tells you when it is done.
- Test retrievalUse the evaluate page to ask the questions your workflow will ask and see what comes back.
- Add a retrieval step to a workflowThe knowledge step pulls the relevant passages into the model step that follows it.
Meets your team where it already works.
One workspace, the rest of the job.
See it on your stack.
30 minutes with our team. We'll walk you through governance, audit, evals — and answer everything procurement will ask. Bring your own NDA; we'll sign in 24 hours.