Upload your documents, attach them to a workflow, and your agents can search them. No vector database to run, no retrieval pipeline to build. Just your knowledge, grounded and cited.
Everything between a folder of documents and a grounded answer, done for you.
Drop in PDFs, docs, and text. Knitch parses and chunks them for you, so they're ready to search in minutes.
Vector and keyword search together, ranked, so it finds the right passage whether the query matches the meaning or the exact words.
Every result names the document it came from, so your agents ground their answers in real text instead of guessing.
Bind a knowledge base to a workflow and its AI nodes can search it. It's the same retrieval the Copilot uses.
Ingestion, embeddings, indexing, and search are handled. There's nothing to provision, tune, or keep running.
See every document in a knowledge base, with chunk counts and last updated, and add or remove docs as your knowledge changes.
The same retrieval your agents and the Copilot use. A knowledge base isn't a separate feature bolted onto one corner of the product. Your AI nodes search it, and the Workflow Copilot searches it, with the same hybrid retrieval underneath.