DocAttaché is a Rovo agent that lets your documents speak. Confluence pages and Jira issues are only part of your team's knowledge — the rest sits in attached files: project reports, budget spreadsheets, slide decks, signed contracts, exported documents. DocAttaché reads those files where they are and brings what's inside them into the conversation, with every answer grounded in the source and linked back to it.
There is nothing to install beyond the app and nothing to configure. DocAttaché is ready the moment it appears in your Rovo agent picker.
Your first answer in 60 seconds
Open any Confluence page — or any Jira issue — that has a file attached. DocAttaché is fluent in PDF, Word, PowerPoint, Excel, OpenDocument text, EPUB, RTF, and CSV.
Open Rovo chat and pick DocAttaché from the agent picker.
Ask about the file. On a page: "What's in the files attached to this page?" On an issue: "What does the attachment on this issue say?"
The answer comes back with every source linked, quoting what is actually in the file. When something isn't in the document, DocAttaché says so plainly rather than guessing.
What you can ask
"What does the attached contract say about payment terms?"
"Summarize the roadmap deck — including the speaker notes." Slide decks often carry their real content in notes that are invisible on the slides; DocAttaché reads those too.
"Which document here mentions the Q3 headcount decision?" — DocAttaché searches inside attachment contents across your Confluence spaces, not just file names.
"Does the attached spreadsheet match the budget on this page?" — it combines the page and the attachment, cites both, and flags any discrepancy.
On Jira: "What does this issue say — description and comments?", or ask it to find related issues by topic or JQL, then read what it finds.
Long documents are no obstacle: DocAttaché reads them in full, section by section, and tells you when it has only read part of a document so far.
Let attachments announce themselves
Atlassian Automation can invoke Rovo agents, DocAttaché included. A rule that fires when an attachment is added, asks DocAttaché to read it, and posts the answer back as a comment means every new attachment can announce what's inside it — hands-free, for the whole team to see.
Build the rule alongside your other automation rules for the space or project; DocAttaché is available there as an agent to call.
Build with it
DocAttaché's skills are available to your own agents in Rovo Agent Studio:
Parse attachment turns an attached file into clean, agent-readable markdown — headings, tables, and structure preserved.
Find attachments searches inside attachment contents across Confluence, not just file names.
Add either skill to a custom agent and it becomes attachment-fluent, without writing a converter.
Private by architecture
Everything happens inside Atlassian's cloud. Document conversion runs in-process in the app — no external converters, no third-party AI services, zero data egress. The Runs on Atlassian badge is the proof, not a promise. The language layer is Rovo itself, under your existing Atlassian AI terms.
Answers respect permissions: DocAttaché reads only what the person asking can already see.
Good to know
Text-based documents only. Scanned or image-only PDFs can't be converted — DocAttaché will tell you when a file can't be read instead of improvising an answer.
Confluence attachment search reads inside files; Jira issue search doesn't. On Confluence, "which document mentions X" searches the indexed text of attachments. On Jira, search covers issue fields — summary, description, comments — and DocAttaché then reads the attachments on the issues it finds.
Jira is optional. DocAttaché installs on Confluence; connect Jira from Atlassian Administration and the same agent reads issues and their attachments too.
Grounded by design. DocAttaché only makes claims about content it has actually retrieved. If it hasn't read something, it says so.
If you see a licensing message, a site admin can start a trial or reactivate the subscription from the Atlassian Marketplace.
