Who we are
Doc Drift is made by Nano-Soft Technology SRL (“Nano Soft”), a software company registered in Romania and based in Bucharest. Nano Soft builds and runs the Doc Drift apps and this website, and is responsible for how they handle data. Website: nano-soft.tech.
The apps run on Atlassian Forge inside your Atlassian site, so your organization stays in control of its Confluence, Jira and Bitbucket data. Atlassian's own handling of that data is covered by Atlassian's terms and privacy policy.
Summary
- No servers of ours. Both apps run on Atlassian Forge. Data stays in Forge storage for your site, and the Bitbucket relay talks only to the Confluence app.
- No page bodies stored. Pages are read to match changes and find affected sections, then discarded. The app keeps anchors, page states, change details and file paths.
- Diff text is deleted after 30 days. The list of changed files and the pull request title stay, so the history still makes sense.
- No AI in detection. AI runs only when someone clicks Propose fix, inside Rovo on your own site.
- Permission-aware. People and agents only see the state of pages they can view.
- No tracking. The apps have no analytics. This website sets no cookies, loads its fonts from Google Fonts, and remembers only your theme choice, in your browser.
Where data lives
Doc Drift is two Forge apps: the main app in Confluence, and a relay in Bitbucket. Forge runs them on Atlassian's cloud. The apps don't send data to any other service.
- App data is kept in Forge hosted storage (Forge SQL and the Key-Value Store). Atlassian encrypts it at rest, keeps each installation's data separate, and stores it in the same location as your Atlassian site, so it follows your data residency settings. Source: Forge storage overview.
- The relay's only outbound address is the Confluence app's Forge web trigger (
*.webtrigger.atlassian.app). Every request is signed with a secret generated on your site. The Confluence app keeps the secret in Forge's secret store and shows only a short fingerprint. - The developer can't browse your data. App logs record IDs, counts and outcomes, and never page text.
What the apps read
| What | Why |
|---|---|
| Confluence pages you anchor | To suggest anchors, match changes to the sections that mention them, and show those sections in the review panel |
| Page owner and your account | To mention the owner in a comment, and to check you can view or edit a page before acting |
| Merged pull requests (relay) | The title, author name, merge commit, list of changed files, and the diff for up to 300 files |
| Anchored Jira issues | When an anchored issue changes: which fields changed, with their old and new values |
| Jira projects you can create in | Only when you open Create Jira task, to list the projects |
What the apps store
| Data | Contents | Kept |
|---|---|---|
| Anchors | Page ID, space key, repository, path or glob, Jira key, weight, who created it | Until removed |
| Page state | State, owner account ID, who verified and when, a one-line change summary | While the page exists |
| Changes | Pull request number, title, author name, merge commit, changed file paths; or a Jira key with the changed fields and values | Kept, for the history |
| Diff text | The raw diff of a merged pull request | 30 days, then deleted |
| Drift events | Which anchor matched, score, severity, reasons, identifiers the change removed, status and who closed it | Kept, for the audit trail |
| Verifications | Who verified a page, when, and an optional note | Kept, for the audit trail |
| Fix drafts | The text of the one block a draft replaces, the proposed text, and who accepted or rejected it | Kept with the decision |
| Settings | Sensitivity, ignored paths, watched spaces, timings, the relay secret | Until changed |
The badge state is also written to a content property on each anchored page, so the badge loads with the page. It holds the state, a title and a tooltip.
When AI is used
Detection never uses AI. A deterministic matcher compares the files and identifiers a change touched with each page's anchors.
AI is used only when a person clicks Propose fix or asks the Drift Reviewer agent in Rovo. Drift Reviewer runs in Rovo on your site, under Atlassian's terms for Rovo, and uses your Rovo credits. It receives the affected sections of the page and a summary of each change with up to 4,000 characters of its diff. It never receives a whole repository. It can save drafts, but only a person can apply one to the page.
The apps call no other AI service.
Who sees what
- The review panel, the agent and the MCP tools read pages as the signed-in person. A page that person can't view comes back as
unknown, with no title. - Anyone who can view a page sees that it drifted and a summary: the change, its author, the file paths and why it matched. Raw diffs and full Jira field values are shown only to people who can edit the page, and never to MCP tools.
- Pages with view restrictions aren't supported yet: the app doesn't watch them.
- Settings are for Confluence administrators. The relay's settings are for Bitbucket workspace admins.
- Coding agents connect through Atlassian's MCP gateway with OAuth, only after an admin turns the tools on for the site, and only as the person who signs in.
What the apps change
- Page text changes only when a person accepts a fix, as that person, with a version message naming the pull request. Page history can undo it.
- Page metadata: the badge content property, written by the app after it checks the clicking person's permission.
- Comments: one footer comment per page that mentions the owner when anchors change. Admins can turn these off.
- Jira tasks are created only when a person clicks Create Jira task, as that person.
Permissions
Doc Drift (Confluence, Jira optional)
| Scope | Used for |
|---|---|
read:page:confluence | Reading anchored pages and their owners |
write:page:confluence | The badge content property, and saving fixes a person accepts |
write:comment:confluence | The owner comment |
read:space:confluence | The space dashboard |
read:confluence-user | Checking who is an administrator |
read:confluence-content.summary, write:confluence-content | Receiving page trashed and deleted events, so their drift closes. Atlassian requires these scopes for those events |
read:jira-work | Changes to anchored Jira issues |
write:jira-work | Create Jira task, only as the person who clicks |
storage:app | Settings and the relay secret |
Doc Drift Relay (Bitbucket)
| Scope | Used for |
|---|---|
read:pullrequest:bitbucket | Merged pull requests and their changed files |
read:repository:bitbucket | The main branch and the diff |
read:workspace:bitbucket | Checking that only workspace admins change the relay settings |
storage:app | The ingest URL, the secret and delivery records |
Retention and deletion
- Diff text is deleted 30 days after a change arrives, by a daily job.
- Removing an anchor deletes it. When a page is trashed or deleted, its anchors and state are deleted and its open drift closes.
- When the app is uninstalled, Atlassian keeps its Forge storage for 28 days so a reinstall can be linked back to it, then deletes it under Atlassian's data retention policy. Source: data lifecycle for Forge-hosted storage.
- Content properties, comments, page versions and Jira tasks the apps created belong to Confluence and Jira, and stay until you delete them there.
To have data for a person or a page removed before then, email hello@nano-soft.tech.
Contact
For privacy questions, data requests or deletion requests, email hello@nano-soft.tech. Tell us your site and the page or person the request is about. Don't send page content or secrets.
Nano-Soft Technology SRL (Nano Soft), Bucharest, Romania. nano-soft.tech
Doc Drift isn't made or endorsed by Atlassian.