Skip to main content

Agents · Knowledge Base

Ground your agents in your knowledge

Give your agent the context it needs. Create knowledge corpora, add documents by text or URL, and let retrieval surface the right passages — so answers are grounded in your own content, not guesswork.

What you can do

A retrieval-ready knowledge base

Bring your documents; the workspace handles ingestion, chunking, and retrieval for you.

Organize into corpora

Group related material into knowledge corpora — one per product, project, or topic — each with its own name and description.

Add documents by text or URL

Paste text or point at a URL. Documents are ingested into the corpus and tracked with status and size.

Chunked automatically

Each document is split into chunks on ingest, so retrieval returns the most relevant passages — not whole files.

Grounded answers

Your agent retrieves from the knowledge base before it answers, so responses are grounded in your own content.

How it works

From documents to grounded answers

  1. 1

    Create a corpus

    Name a knowledge base for a product, project, or topic.

  2. 2

    Add documents

    Ingest content by pasting text or linking a URL.

  3. 3

    Auto-chunk

    Documents are chunked and indexed for retrieval.

  4. 4

    Ground answers

    Your agent retrieves the right context when it responds.

Give your agent a memory of your docs

Create a corpus, add your content, and your agent starts answering from it.