Skip to main content
A Knowledge Base is a repository of documents, FAQs, and information that your agent can search and reference during conversations. Instead of trying to stuff everything into a prompt, you upload content and let the agent retrieve relevant information when needed.

Why Use Knowledge Base

More Accurate Answers

Agents can cite actual documents rather than guessing or hallucinating.

Easy Updates

Change a policy? Update the document. No prompt editing needed.

Scalable Knowledge

Add hundreds of pages of content that would never fit in a prompt.

Reduced Prompt Complexity

Keep your prompt focused on behavior. Let the KB handle information.

How It Works

  1. You upload content — Documents, text, URLs
  2. Atoms processes it — Content is indexed for search
  3. Caller asks a question — Agent receives the query
  4. Agent searches KB — Finds relevant sections
  5. Agent responds — Uses KB content in the answer

What You Can Upload

Content TypeExamples
DocumentsPDFs, Word docs, text files
TextDirectly pasted content
URLsWeb pages to crawl
FAQ listsQuestion-answer pairs
→ See: Uploading Content for details

Local vs Global

TypeScopeUse Case
Local KBSingle agentAgent-specific information
Global KBMultiple agentsShared company knowledge
Example:
  • Global KB: Company policies, general FAQ
  • Local KB: Product-specific details for a product support agent
→ See: Local vs Global

Use Cases

Customer Support

  • Product documentation
  • Troubleshooting guides
  • Return/refund policies
  • Shipping information

Sales

  • Product specifications
  • Pricing sheets
  • Competitive comparisons
  • Case studies

Healthcare

  • Appointment policies
  • Insurance information
  • Procedure descriptions
  • Location details

Any Industry

  • FAQs
  • Policies
  • Product information
  • How-to guides

Getting Started

  1. Create a Knowledge Base — Give it a name and type
  2. Upload content — Add your documents and information
  3. Attach to agent — Connect KB in Configuration Panel
  4. Test — Verify agent finds and uses the right information

What’s Next