Knowledge Integrations — Confluence, SharePoint, Notion

Most companies already have a wiki. Don’t rebuild your knowledge base inside Omniflow — sync the wiki you have. Sync runs on a schedule, embeddings stay fresh, and your AI agents answer from the same articles your humans read.

Supported sources

Source	Auth	Sync frequency	Format support
Atlassian Confluence	OAuth	Hourly	Pages, attachments, labels
Microsoft SharePoint	OAuth	Hourly	Pages, documents, sites
Notion	OAuth	Hourly	Pages, databases
Google Drive	OAuth	Hourly	Docs, Slides, PDFs
Public URL crawl	None	Daily	HTML, sitemaps
GitHub repo (markdown)	OAuth	Hourly	`.md`, `.mdx`

Connect Confluence

Authorize

Settings → Integrations → Confluence → Connect. The connector requests read:space, read:page, and read:attachment.

Pick spaces

You almost never want to sync everything. Pick the customer-facing spaces and the internal SOPs you want agents to see.

Pick labels (optional)

Filter by label — support-public, agent-runbook, etc. Pages without one of the picked labels are skipped.

Decide what to embed

Setting	Default
Page bodies	Yes
Attached PDFs	Yes
Page comments	No
Archived pages	No

Save and run initial sync

Initial sync embeds every selected page. For a 5,000-page space, this takes 30-60 minutes. Subsequent syncs are incremental.

Connect SharePoint

The flow is similar but with site/library granularity:

Authorize with a user that has read access to the libraries you want.
Pick sites and libraries.
Choose document types (Pages, Word, PDF).
Run.

SharePoint sync is sensitive to permissions — if a user can’t read a library, the integration can’t see it.

Notion

Notion uses page-level permissions. The integration sees pages explicitly shared with it (via the Notion app). Share at the parent page level so child pages inherit access.

Public URL crawl

For knowledge bases that aren’t on a wiki — your help center on Zendesk, Intercom, or a custom CMS:

Add the root URL.
Optionally provide a sitemap URL for faster discovery.
Set crawl depth and include/exclude patterns.

The crawler runs daily and surfaces diffs in the Knowledge → Activity log.

Using synced knowledge

Synced articles appear in the Omniflow knowledge browser alongside Omniflow-authored articles. Attach them to agents in the Knowledge tab the same way:

Whole spaces / libraries / sites.
Specific labels.
Specific articles.

When the agent retrieves a chunk, the trace links back to the source URL — agents (and customers, if you cite) can click through to the original.

Permissions are honored. If a synced article is restricted to a Confluence space the agent shouldn’t see, scope the integration to exclude it. Don’t rely on document-level ACLs at retrieval time — the integration enforces access at sync.

Keeping knowledge fresh

Event	What happens
Article edited in Confluence	Re-embedded on the next sync (typically <1h).
Article deleted	Removed from the index on the next sync.
Permissions tightened	The article disappears from sync once the integration user loses access.
Sync paused	Existing index stays; nothing is updated until you resume.

Pitfalls

Pitfall	Fix
Outdated page lingering	Confirm the page was deleted in source; force sync.
Massive PDF blowing token budget	Exclude attachments above a size threshold.
Internal-only content leaking to a public agent	Scope the integration tightly; double-check via test calls.

Wiki content tends to drift faster than it gets reviewed. Pair the integration with a quarterly content audit so stale articles don’t poison agent answers.

Open in Omniflow

If you want to…	Go to
Connect knowledge to an agent	Knowledge Sources
Author articles in Omniflow	Knowledge Base
Inspect retrieval	Activity Logs & Traces