Knowledge scope
Control exactly which records and knowledge bases your agent can read and reason about, so it stays focused, private, and efficient.
Knowledge scope controls which records and knowledge bases an agent is allowed to see. You set it in the Knowledge tab on the agent detail page.
Scoping matters for three reasons: privacy (the agent only sees what it should), focus (it reasons over the right data), and cost (less irrelevant context to process).
How scoping works
The Knowledge tab shows a tree of your resource types. For each, you choose how the agent may access it:
| Mode | Meaning |
|---|---|
| Include | The agent can read this record type, or a single record |
| Include with children | The agent can read this record and everything beneath it |
| Exclude | The agent explicitly cannot read this — exclusions always win |
You can scope at two levels:
- A whole type — for example, all tickets or all contacts.
- A single record — just one specific contact, ticket, or article.
Resource types you can scope
Scoping covers system resources and your own custom data:
- Knowledge Bases
- Contacts and Companies
- Tickets
- Datasets
- Any custom entities you've defined
Pinning vs. mentioning
There are two ways to add a record to the scope:
- Pin it in the tree — a normal grant you can remove anytime.
- Mention it in the persona prompt — this locks it into the scope until you remove the mention, the same way tool mentions work.
Exclusions take priority
If a record is both included by a broad grant and excluded specifically, the exclusion wins. Use this to carve out exceptions from a wide grant.