Skip to main content
The Company Brain answers from two kinds of content:
  • Collections: files, pasted text and web pages stored in xpander, searched by meaning and by exact words.
  • Live sources: Google Drive, SharePoint, OneDrive, Confluence, Jira, Notion and Salesforce, searched at question time with the account of the person asking. Nothing is copied into xpander unless an admin keeps a copy of a folder. Your own systems join as custom sources, over MCP or REST. On an Air-Gapped install, live sources are custom sources only (Network).
Every agent asks the Company Brain by default. Admins can also build specific brains, such as Legal or Support, and attach them to the agents that need them. Admins set it up in Platform > Agents > Knowledge.
On every live source an admin can also exclude by path, label, file type and age. Uploads can be PDF, Word, PowerPoint, Excel, Markdown, HTML or images, up to 25 MB each. Scanned pages are read with OCR.

Each person gets only what they can already open

  • Stored documents are filtered by who may read them before ranking, so a document someone may not open never reaches the answer.
  • Live sources are read with the person’s own account. A permission removed in the source applies on the next question.
  • Sources a person has not connected give them nothing. There is no shared or service account to fall back on. The answer names the skipped sources and offers Connect for each.
A specific brain only narrows where an agent looks. The one exception is a brain set to Shared agent access, for public agents whose visitors have no company login. See Whose access it uses.

Citations in the conversation

An answer that used the brain ends with Sources: numbered chips that match the numbers in the text.
  • A stored document opens over the answer with the cited passage highlighted. Admins also get Manage in Knowledge.
  • A live item opens where it lives, with your own account.
Sources list with 1 Enterprise MSA Playbook v4 (Collection, Legal playbooks, p. 7), 2 Deal Desk Exceptions (Confluence, REVOPS) and 3 Q3 Contract Review minutes (SharePoint, Legal)

The Sources row under an answer, opened: a stored document with its collection and page, and two live items with their source. Shown with sample data.

Enterprise MSA Playbook dialog, Collection, Legal playbooks, p. 7, listing passages with the cited section 9 Liability, 9.2 Cap highlighted

A cited collection document opened from the answer, with the cited passage highlighted. Shown with sample data.

Your knowledge

Open Settings > Knowledge in xpander Chat. Nothing on it needs an admin.
Settings, Knowledge tab with My sources (Google Drive and Jira with Connect, SharePoint and Confluence connected with Disconnect), Shared with you, My collections with New collection and Add files, and Waiting for your approval (2)

Settings > Knowledge: your own sign-ins, the collections shared with you, your own collections, and facts waiting for your approval. Shown with sample data.

  • My sources: Connect opens the provider’s sign-in, once per source. Disconnect stops the brain reading that source for you at once. You can also connect from Connect under an answer that skipped a source.
  • Shared with you: every collection you can read. When the owner lets readers add files, the row offers Add files.
  • My collections: your own collections, when an admin turns on Members can create their own collections. Select New collection, fill Name and What it holds, pick who can read it, and add files.
  • Waiting for your approval: facts agents learned from your conversations. A fact is used only after you Approve it.
My sources lists the live sources your admins set up. Agents read them with your own account, so they find only what you can open, and a source you have not connected is skipped. A row that says why it is not connected, such as Not connected or Sign-in refused, offers Connect. No Connect means the source is not fully set up: ask an admin. Notion signs you in with your own Notion account; there are no pages to pick or share, and Disconnect also signs you out of Notion for xpander. After you connect from a conversation, your next question reads the source as you.Shared with you holds every collection you can read, other than your own. Agents answer you only from files you may open. New files are ready to answer from in a minute or two.My collections, for notes, playbooks or team files:
  1. Select New collection, then fill Name and What it holds.
  2. Under Who can read it, pick people and groups. A new collection is yours alone until you share it. Everyone in the organization appears only when your admin allows it.
  3. Turn on Let others who can read it add files if your team should add to it.
  4. Select Create collection, then upload files, paste text, or import a public web page.
The collection menu offers Rename and share, Add files and Delete collection. Deleting removes its files from every answer.Waiting for your approval: Edit changes the text first (Save and approve), and Reject discards it. Facts proposed for a collection or brain you own come here too. See Learned facts.

Attach knowledge to an agent

Agent settings > KNOWLEDGE lists what one agent can look up:
  • Company Brain: on by default for every agent. Turn it off to keep this agent away from company knowledge.
  • Attached brains, each marked “each person’s access” or “shared agent access”.
  • Attached collections. A collection attached directly always uses each person’s own access.
Add knowledge attaches a brain or collection.
Knowledge row with the Company Brain switched on and the Add knowledge menu listing the Legal, Support KB and Pricing brains and four collections with their file counts

Add knowledge on an internal agent: the Company Brain is on, and the menu lists the brains and collections you can attach. Shown with sample data.

On a public agent, visitors have no company login. The Company Brain shows “Off. On a public agent it would be empty for visitors without a company login.”, and brains and collections that use each person’s own access find nothing there. Attach a brain with shared agent access instead, which holds only collections. To include files from Google Drive, SharePoint or OneDrive, an admin first keeps a copy of the folder in a collection.
Agent settings with the Knowledge row: Company Brain switched off for a public agent, the Support KB and Pricing brains using shared agent access, the Runbooks collection, and Add knowledge

The Knowledge row on a public agent: the Company Brain is off, and two brains with shared agent access and a collection are attached. Shown with sample data.

When Add knowledge offers nothing, admins can create a collection or connect a source, and members can make their own collection in Settings > Knowledge.The agent asks all its brains at once: the Company Brain, when on, and the attached brains. It can also ask one brain by name, chosen from the brain’s What it’s for description. An agent can ask only brains attached to it. Live sources are searched in parallel. A source that does not answer in time is left out of that answer and named. When a passage is too short, the agent opens the cited item in full.

Ask the brain from an MCP client

MCP clients connected to xpander, such as Claude Code, Claude Desktop, ChatGPT, Codex, Cursor and OpenCode, can ask the same brain. See Connect your MCP client. Every call runs as the person who signed in. For example, in Claude Code:
Learned facts are proposed only from agent conversations. MCP clients can ask and read, not propose.

Troubleshooting

Check, in order:
  1. Can you open it yourself? The brain reads as you, so it finds only what your own account can open.
  2. Is the source connected for you? Open Settings > Knowledge > My sources. A source you never connected is skipped, and the answer offers Connect.
  3. Is it inside the source’s scope? An admin can run Test as a user on the source card with your name and the same question. The result lists what was dropped by the scope and why, for example “not on the allowlist” or “older than the cut-off”.
  4. Is the source on this agent? A specific brain may hold only part of a source, and the Company Brain may be switched off for the agent under Agent settings > KNOWLEDGE.
  5. Is the document ready? A stored document is searched only once its status is Ready.
The source was not read for this question. Usually you have not connected your account (Connect under the answer fixes it), the provider refused your sign-in (reconnect from Settings > Knowledge), or the source did not answer in time. The rest of the answer still stands. Ask again to retry the source.
The source is not fully set up yet. An admin opens Platform > Agents > Knowledge, where the Needs you strip lists sources that are not set up.
  • Collection audience or document access in xpander: on the next question. Nothing is parsed again.
  • Permission removed in a live source: on the next question once the person’s cached results expire, at most the source’s Cache results for time (60 seconds by default). Sensitive sources are never cached.
  • A person disconnects or their sign-in is revoked: their cached results for that source are cleared at once.
No. The brain reads live sources with the account of the person asking, and the Company Brain always reads as that person. Admins choose which sources and collections exist and what a source’s scope allows. They never get anyone else’s access. Test as a user shows titles and the reason each item would be read or left out by the scope, not the content.
No, unless an admin keeps a copy of a folder. Live results are cached briefly per person, encrypted, and expire on their own. Sensitive sources are never cached. The source card’s Connection tab says whether content is stored.