Skip to content

Knowledge

Knowledge sources provide structured information that AI agents can reference during conversations — giving your workflows access to data beyond their training.

Knowledge gives your workflows a reusable place to store reference data that should stay available across many conversations. Instead of teaching an AI agent the same facts repeatedly in prompts, you can keep those facts in a structured surface the workflow can search and reuse.

  • Use knowledge when an AI agent should reference structured facts during a conversation.
  • Use Sheets when that information is table-shaped and managed inside Octo.
  • Use Tools when the workflow needs live data from another system at runtime.

Knowledge works best for information that is stable or changes on an operational cadence, not on every request. Tools are better for real-time lookups, transactions, and system-of-record actions.

Sheets are a good fit when the workflow needs:

  • policy or eligibility tables,
  • product or plan catalogs,
  • routing or escalation rules,
  • approved response snippets,
  • operating hours or location data,
  • or any other structured facts that are easier to maintain as rows and columns.

Choose Tools instead when the workflow should:

  • fetch live account or order data,
  • create or update records in another system,
  • send a message or trigger an external action,
  • or rely on information that should always come from an API at conversation time.

Before connecting knowledge to a workflow:

  • keep one concept per row,
  • use clear header names,
  • add descriptions for ambiguous columns,
  • test the data shape with real retrieval questions,
  • and avoid storing secrets or credentials in the sheet.

Those small decisions make the difference between a sheet that looks good to a human and one that is actually useful to an AI workflow.

The main knowledge surface in Octo is Sheets, a spreadsheet-like editor for tabular reference data with searchable rows, column descriptions, and SQL querying.

  • Open Sheets to create and manage knowledge tables.
  • Use AI Workflows when you are ready to connect knowledge to agent behavior.
  • Use Tools when the workflow needs live data instead of reference data.
  • AI Workflows — workflows reference knowledge sources during execution
  • Tools — call live systems when sheet data is not enough