Does Rovo's chat have any governance for information sources?

[Andrea Del Pilar Porras Rodríguez]

I would like to know if Rovo Chat allows for the configuration of knowledge sources and if it is possible to set limits on them.

[Rebekka Heilmann _viadee_]

Hi @Andrea Del Pilar Porras Rodríguez - welcome to the Community,

plain Rovo Chat can be configured via the icons on the bottom. You can uncheck connected Apps there. You cannot (de-)select single Spaces. You can however specify what you want to search in the prompt. This is not a garantuee that Rovo is not looking elsewhere but makes it much less likely.

Like @Ajay _view26_ explained, Rovo Agents have specific access to knowledge based on their configuration. When you chat with Agents, you can not limit this any further.

Rovo respects User permissions and will not show information the user doesn't have access to. Atlassian announced Agent Accounts that will have their own permissions - so there may be some Agents in the future where user permissions don't matter, but this is all down to the specific configurations.

Apart from the App Blocklist, there is no other option to restrict certain spaces from showing up in Rovo in general. It's always down to specific Chat/Agent setup and permissions.

[Ajay _view26_]

Hi @Andrea Del Pilar Porras Rodríguez 

Welcome to the community!

Good question — the short answer is yes, but the controls sit at different layers depending on whether you're talking about Rovo Chat itself or Rovo Agents.

By default, Chat pulls from everything in your Teamwork Graph — all connected Atlassian products plus any third-party connectors your org admin has set up (Google Drive, Slack, SharePoint, etc.). There's no way for individual users to restrict which sources Chat queries during a conversation. However, Chat always respects existing permissions, so users only ever see content they already have access to in the source systems.

In the case of custom agent, each subagent can be configured with one of three knowledge scope options: "All organizational knowledge" (the default — everything connected), "Custom knowledge" (you pick specific Confluence spaces, Jira projects, Google Drive folders, etc.), or "No organizational knowledge" (the agent only uses its instructions and optionally web search). This is where you can meaningfully restrict what sources the AI draws from for a given use case.