> ## Documentation Index
> Fetch the complete documentation index at: https://docs.letsflow.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Knowledge

> Knowledge hub: articles and the sources that feed grounded answers.

Knowledge covers how Flow ingests articles from connected sources, searches them at runtime, respects
source permissions, and uses team or organization instructions to shape its answers.

## Knowledge hub

The sidebar label **Knowledge** opens the knowledge hub. Flow combines:

* **Ingested articles** from connected sources (Confluence, SharePoint, Google Drive, ServiceNow KB, and others)
* **Search** across those articles at runtime

The hub lists articles. Knowledge sources are connected and managed under
**Integrations**, filtered to knowledge integrations. Sync status and source
issues appear on each source there. Team Flow instructions are on the **Flow instructions**
tab in **Settings → Flow Teams**. Organization instructions are in **Settings → Global AI settings**.

## Connect a knowledge source

Administrators and knowledge managers:

<Steps>
  <Step title="Open knowledge integrations.">
    Open **Integrations** and filter to knowledge integrations.
  </Step>

  <Step title="Start the connection.">
    Choose the provider, then add a source to start the connection flow.
  </Step>

  <Step title="Select the scope.">
    Select the team and content scope, such as a site, library, space, folder, or knowledge base.
  </Step>

  <Step title="Sync the source.">
    Trigger or wait for sync. New articles appear under **Knowledge** when ingestion completes.
  </Step>
</Steps>

Teams need knowledge enabled under team settings.

A team connection is where you choose the source and watch sync health. It is
not the read lock. For SharePoint, SharePoint Scoped, Google Drive, Dropbox,
Box, and ServiceNow, Flow can mirror source permissions. In that mode, the
source must already grant a person access, and Flow must have stored that
grant. Connecting the source to their team is not required to read. Connecting
it does not grant read.
Organization admins cannot bypass that source access.

When Flow cannot determine who the source allows, it does not grant everyone
access. Some SharePoint pages stay limited to the team that connected the
source. ServiceNow can use an explicit team or organization setting. See
[SharePoint](/integrations/knowledge/sharepoint) and
[ServiceNow KB](/integrations/knowledge/servicenow-kb).

Confluence Cloud can mirror those Confluence permissions, or an administrator
can choose a Flow publication audience: the teams that connected the source, or
the organization. Publication-audience connections skip restricted pages. See
[Confluence Cloud](/integrations/knowledge/confluence).

Notion is different. Flow does not copy Notion per-person sharing. An
administrator chooses a Flow publication audience instead. See
[Knowledge integrations](/integrations/knowledge/overview).

## Source status

| Label | Meaning |
| - | - |
| **Connected** | The provider account and selected scope are connected. This does not confirm freshness. |
| **Indexed** | Files currently available to Flow from that source. |
| **Not synced yet** | Flow has no completed sync time to show. Run or wait for the first sync before relying on freshness. |
| **Last successful sync** | The most recent completed refresh. |

Use the source's actions menu to sync or reconnect when those actions are available. If **Indexed**
shows files while **Sync** says **Not synced yet**, treat the files as an existing snapshot and run
the first tracked sync before relying on them.

A source can finish a sync and still show a warning. For ServiceNow Knowledge, **ServiceNow count
mismatch** means the provider count was higher than the articles returned. The sync completed. See
[ServiceNow Knowledge Base](/integrations/knowledge/servicenow-kb).

## Find an article

<Steps>
  <Step title="Open Knowledge.">
    Open **Knowledge** from the sidebar.
  </Step>

  <Step title="Find the document.">
    Use the list to find documents. The source shows where each article came from.
  </Step>

  <Step title="Open Document detail.">
    Select a row to open **Document detail**:

    * Content preview
    * Sync status
    * Source metadata
  </Step>
</Steps>

The org-wide **Knowledge** list shows files the person can already open in the
source, after Flow has stored that grant. Team-scoped views also require a
source that team owns or later connected.

Use articles to validate ingestion before relying on AI answers in production.

## Configure Flow instructions

<Steps>
  <Step title="Open the team.">
    Open **Settings → Flow Teams** and expand the team.
  </Step>

  <Step title="Open Flow instructions.">
    Open the **Flow instructions** tab.
  </Step>

  <Step title="Open organization instructions.">
    Open **Settings → Global AI settings**.
  </Step>

  <Step title="Edit and save.">
    Edit and save the draft.
  </Step>

  <Step title="Publish.">
    Publish when it is ready. A saved draft is not active.
  </Step>
</Steps>

Team instructions are limited to **15,968 characters** so they fit within Flow's runtime customer-prompt budget (4,000 tokens). Content over that limit cannot be saved or published.

Organization admins and team managers can edit team Flow instructions.

## Searching from Home

The **Home** prompt bar accepts natural-language questions. Flow may use synced knowledge and
configured sources such as linked Jira issues. Org-wide search uses the same source
access as the **Knowledge** list. It does not require the person's team to
connect the source. Team-scoped search still limits results to sources that
team owns or connected, then applies the same source access.

When search contributes evidence, Flow can show source cards beneath the answer. Each card links to
the original document in its connected source, not to **Document detail** in the **Knowledge** hub.
Open them when the answer affects a decision. If content may be stale, check the source's latest
successful sync under **Integrations**. Flow may use other authorized context when no eligible
source is available, but it does not treat an undocumented company policy as fact.

Federated options depend on your connected integrations. See [Search](/integrations/knowledge/search).

## When Flow cannot confirm an answer

Flow does not automatically know every company policy. If a policy is not documented, is outside
your access, or is missing from connected sources, Flow cannot confirm the answer. Ask the policy
owner for clarification, then document the decision for future questions.

## Who does what

| Role | Typical tasks |
| - | - |
| **Administrator** | Connect providers, org-level policies |
| **Knowledge manager** | Curate sources, verify article sync |
| **Team admin** | Enable knowledge for the team, team Flow instructions |
| **Operator** | Consume answers via Flow and ticket AI thread |

## Related topics

* [Application access](/setup/applications)
* [Integrations: Knowledge](/integrations/knowledge/overview)
* [Help desk](/setup/help-desk): using AI thread on conversations


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.