Enclave finds sensitive data in your Confluence Cloud — personal information, credentials, and financial data — across pages and attachments, classifies pages by sensitivity, and lets you redact or mask what it finds, with a full audit trail. It runs entirely inside your own Atlassian Cloud instance: your content is never sent to any external server.
Enclave is built on Atlassian Forge and runs inside your Atlassian Cloud tenant. When it scans a page or attachment, the content is read and analyzed in place — it is never transmitted to the developer or to any third-party service. The app declares no external network access.
The app stores only masked previews (for example, the last four digits of a card number) and metadata (what type of data was found, how many, and where), using Atlassian’s own storage within your instance. Raw sensitive values are never stored.
Install Enclave from the Atlassian Marketplace. After installation, find it under Apps → Enclave. New installations begin a free trial; see Licensing.
That’s the core loop: scan → review → act. Everything below adds control and depth.
Detectors identify specific kinds of sensitive data (emails, credit cards, API keys, national ID numbers, and so on). Detectors are grouped into packs you enable by relevance and jurisdiction.
Confidence tiers. Each finding carries a confidence level:
This tiering is what keeps automatic actions trustworthy: only high-confidence findings drive automatic classification.
Sensitivity levels describe how sensitive a page is — for example Public, Internal, Confidential, Restricted. A page’s automatic level is driven by the most sensitive thing found on it.
Manual scan. Choose a space and select Start scan. Scans run asynchronously and scale to large spaces.
Scheduled scan. Configure a recurring (for example daily) scan of a chosen space so compliance monitoring is continuous rather than a one-time snapshot. Scheduled scanning is off until you set a target space.
Each finding shows the page, the location (page body, title, or a named attachment), the detector type, the confidence tier, a masked preview, and any classification or remediation status. Not-scanned items are shown too — if the app couldn’t scan something (an oversized attachment, an image-only PDF, an unreadable file), it tells you, so gaps are visible rather than silent.
Enclave scans the text of these attachment types in place, without the file leaving your instance:
.docx).xlsx).txt)Attachment limits (by design):
Enclave can classify pages by sensitivity and apply a matching label you can search and report on.
From a finding you can take one of three actions. Every action requires confirmation and is recorded in the audit log.
Before any redaction or mask, you see a preview of the exact change, with the value masked. After the change is applied, the app re-checks the page to confirm the value was actually removed before reporting success — if it can’t fully remove a value, it tells you rather than claiming success.
Important notes:
Packs let you enable detectors by relevance and jurisdiction, so you only see what’s relevant to you. Typical packs include contact data, financial data, secrets/credentials, and national-identifier packs for specific regions. Enable packs in Settings.
Custom rules let you detect your own formats (for example, an internal ID scheme) using a regular expression. Custom rules run on a safe, non-backtracking engine and can be tested against sample text before you save them. Because a custom pattern isn’t independently validated, custom-rule findings default to a lower confidence and require a context hint to be treated as higher confidence.
Policies let you define where to look, what to match, and what to do — across your spaces. A policy combines a scope (spaces), a match (detectors, packs, confidence, or custom rules), and an action (classify, alert, notify, or queue for remediation).
From the Reports tab you can generate:
All reports contain masked previews only — never raw sensitive values. Large reports generate in the background.
Enclave requests only the Confluence permissions it needs to read content, read and scan attachments, apply labels and classifications, and (for remediation) update pages. Each requested permission is justified at listing time.
Zero egress. The app declares no external network access. To read an attachment’s bytes, it follows Confluence’s own redirect to Atlassian’s media storage — validated against an allowlist of Atlassian hosts, and skipped if the destination isn’t an Atlassian host. Your content is scanned within Atlassian’s platform and is never sent anywhere else.
When unlicensed, scanning and related features are unavailable and the app shows an upgrade prompt; your existing stored findings and settings are not deleted.
We list these plainly so you can rely on the app with clear expectations:
A scan found nothing on a page I expected it to flag. Confirm the relevant detector pack is enabled for your region in Settings, and that the data is structured (see Limitations). Free-prose personal data isn’t detected in this version.
A finding looks like a false positive. Check its confidence tier. Low/medium findings are informational or review-only and never auto-classify. If a high-confidence detector is noisy for your content, you can disable it or its pack in Settings.
An attachment shows as “not scanned.” It’s likely oversized, an image-only/scanned PDF, or an unsupported type — the reason is shown with the item.
Scanning is unavailable / I see an upgrade prompt. Your trial may have ended or the app is unlicensed. See Licensing.
Where did my data go after I uninstalled? Findings and settings are stored within your Atlassian environment and are removed according to Atlassian’s handling of uninstalled apps.
Contact enclave.app.atlassian@gmail.com. When contacting support, please do not include raw sensitive data (such as full card or ID numbers) in your message.