TopolearnAgents only know what they were trained on; Brain grounds their answers in your own, current documents.
Ask an AI model about your holiday policy and it will guess. Brain is Gumloop's company knowledge base: you connect sources, Gumloop reads and indexes them (builds a searchable catalogue), and agents search them and answer with citations (links to the documents they used), so you can check the answer.
Sources include Google Drive, Notion, Slack, Confluence, SharePoint, GitHub, Zendesk, Salesforce, Gmail, Linear and uploaded files. You choose exactly which folders, channels or labels to include. For other apps, an AI setup chat proposes what to sync and waits for your approval.
Each source has a scope:
- Personal: only you.
- Team: members of one team.
- Organization: everyone, or selected teams.
For several sources you can also choose document access: "Original source" mirrors each document's sharing permissions, so people only find what they could open anyway. If Brain can't confirm access, it hides the document.
Brain re-syncs most sources automatically; uploaded files don't update by themselves. Images aren't read as text, and private Slack channels aren't synced.
To use it, attach sources under an agent's Knowledge Sources, optionally narrowed to certain folders. The agent then gets a search tool and a document reader, and decides when to search. Tell it in its instructions to prefer Brain for company questions and to say so when nothing relevant is found.
Indexing and searches use credits.
Remember
- Brain indexes your sources so agents can search and cite them
- Scopes (personal, team, organization) and document access control who sees what
- Attach sources per agent and tell it when to use them
Watch out for
- Uploading a policy PDF once and forgetting it won't update when the policy changes.