Skip to main content
The Confluence integration ingests your team’s SOPs, runbooks, and other operational documentation as a knowledge source for LuumenAI. Once connected, the AI can ground its command recommendations in your team’s actual procedures, suggest reusable Skills when it spots a pattern worth automating, and flag when day-to-day operations are drifting from what the SOP says.

​
What it does

SOP ingestion

Pull pages from selected Confluence spaces into Luumen as searchable knowledge. New and updated pages are picked up continuously.

SOP-aware AI

When you ask LuumenAI for help on a host, it consults your SOPs first. Suggested commands and procedures come from your own documentation, with citations.

Skill recommendations

When the AI notices a procedure your team runs repeatedly, it proposes converting it into a Skill — a reusable automation that captures the runbook in executable form.

SOP drift detection

When operators deviate from the documented procedure, Luumen records the deviation and (with your approval) suggests an update to the Confluence page so the SOP stays current.

​
How knowledge ingestion works

1

Select spaces

During onboarding you pick the Confluence spaces Luumen should index. Most customers start with one or two operations-focused spaces (e.g., “Infrastructure Runbooks”, “Incident Response”) rather than the whole site.
2

Initial sync

Luumen pulls every page in scope, processes the content (Markdown, tables, embedded code), and indexes it for semantic search.
3

Continuous updates

New pages, edits, and deletions in the selected spaces are picked up continuously. The index reflects the current state of your documentation, not a stale snapshot.
4

LuumenAI consults the index

When you start a chat with LuumenAI, it has access to the indexed SOPs as context. Recommendations cite the page they came from so you can verify and follow up.

​
Skill recommendations

Skills are reusable AI workflows built from your SOPs. The Confluence integration is the primary on-ramp for them: when the AI sees a procedure your team is running by hand repeatedly — and an SOP exists for it — it can propose turning that SOP into a Skill. The recommendation flow:
  1. AI detects a repeated pattern (same SOP cited across multiple recent chats, or a procedure that comes up frequently in SNOW incidents).
  2. AI proposes a Skill: a name, the trigger conditions, and the steps drawn from the SOP.
  3. You review the proposal, edit if needed, and approve it.
  4. The Skill becomes available across your workspace. Future invocations follow the same human-in-the-loop pattern as any other Skill.
You don’t have to wait for a recommendation. You can also browse your indexed SOPs in the Luumen UI and explicitly say “turn this page into a Skill” on any of them.

​
SOP drift detection

Operational reality and documented procedure can diverge. When that happens, Luumen tries to catch it.
  • During an AI chat, if you approve commands that materially differ from the cited SOP — different order, different tool, additional verification step — Luumen records the deviation.
  • In a periodic review, those deviations are aggregated. If a pattern emerges (most operators are doing X, but the SOP says Y), Luumen surfaces it.
  • With your approval, Luumen can draft an update to the Confluence page reflecting the actual current practice. The draft is a proposed edit, not an automatic write — you review and approve before it commits to Confluence.
This keeps SOPs current without manual review cycles, while keeping a human in the loop on every change to documentation.

​
What you’ll set up

Connection setup happens during onboarding. The basics:
  • A Confluence service account with read access to the spaces you want indexed (and edit access if you want Luumen to propose drift-detected updates).
  • An API token or OAuth app for the service account.
  • The list of spaces to ingest. You can adjust this later from workspace settings.
Once configured, the integration appears in your workspace settings and the indexed pages become visible to LuumenAI immediately on completion of the initial sync.

​
Audit and control

  • All writes are explicit. Luumen never edits a Confluence page without your approval. Drift-detected updates are presented as proposals.
  • Workspace-scoped indexing. Each Luumen workspace indexes its own configured spaces. Documentation isn’t shared across workspaces or customers.
  • AI data opt-out. If you’ve opted out of AI data usage for your workspace, your SOP content is used only for direct retrieval, not for training or model improvement. See Integrations overview.

​
Limits

  • Confluence Cloud and Data Center are both supported, but the setup details differ. Apiphani engineers will walk you through whichever applies.
  • Page-level granularity. Luumen indexes whole pages. Section-level scoping (only this part of this page) is not supported today.
  • Attachments are not indexed. PDFs, images, and other attachments embedded in Confluence pages are skipped. Inline content (text, code, tables) is indexed.
  • English-language indexing. Like ServiceNow, the semantic index is tuned for English content today.