Atlassian Rovo: Search, Chat, Agents and Studio

Atlassian Rovo can help your team find the decision behind a Jira work item, turn Confluence knowledge into a useful answer, and hand repeatable work to an agent. For a team already using both apps, the strongest starting point is a recurring handover: gather the evidence, draft the update, then have its owner check it.

As of November 2026, the useful distinction is Search finds, Chat explains, Agents carry out a defined job, and Studio builds the agents and workflows. Rovo AI is the broader experience, not just the chat panel.

Atlassian's inclusion line: “Rovo is included in select Enterprise, Premium and Standard Cloud plans, as well as Teamwork Collection.” Rovo features and availability.

What Rovo's Parts Do Inside Jira and Confluence

Rovo brings discovery, conversation and repeatable work into the places your team already uses. Think of it as a company library with a librarian, specialist assistants and a workshop for configuring them.

Search, Chat and Agents use the accessible information connected to your organization. They do not give every employee the same view of company knowledge. Rovo's product overview.

Chat can also take actions, including creating a Jira work item or publishing a Confluence page. An agent makes a particular role and its instructions reusable; action-taking is not exclusive to agents. Chat actions and confirmation.

Studio is now a standalone builder app. Its current workspace covers agents, automation flows and no-code apps. Open it from the app switcher. Atlassian calls the context behind these experiences the Teamwork Graph: connections between people, work and content. Picture the cross-references in a library catalogue, linking a decision to its project and participants. Current Studio capabilities.

For ordinary work, start with a specific question. On a Jira work item, try: “Explain this work item's goal and identify unanswered questions.” On a Confluence proposal, try: “Review this page against the linked decision criteria and show where evidence is missing.” Include the relevant links and time period. Chat can use the content you are viewing and context you explicitly add. Chat context and prompting options.

The business consequence is a shorter preparation chain. A release lead normally collects work items, finds decisions, reconciles them and writes an update. A useful Rovo workflow produces a first draft with evidence for the lead to reconcile. Measure the total hours including checking and corrections; fast generation alone is not a saving.

Outside-App Connectors: Check How the Data Arrives

Connecting an app does not always mean Rovo has indexed its whole workspace. Atlassian now calls these Teamwork Graph connectors, with three access patterns.

Synced and Direct connectors require admin setup. Smart Link connectors do not. Some supported synced connectors can use synchronized permissions without each person signing in separately; others require individual authentication. Connector types and their requirements.

Useful outside sources include documents in Google Drive or Microsoft SharePoint, discussions in Slack, and repositories in GitHub. Figma designs and FigJams also appear in the Smart Link catalogue. Check each connector's supported content types before promising coverage.

There is another distinction in the directory: Included connectors can supply default tools that users authenticate to run, while Added connectors are configured by an admin for broader retrieval or indexing. Seeing an included tool is not proof that an app's content is fully searchable. Supported connectors and setup paths.

For week one, connect the repository that contains a necessary missing input. Adding every available connector makes failures harder to diagnose. The same pilot discipline applies to broader AI enterprise search deployments.

Build Your First Rovo Agent in Week One

Start with a Release Brief agent that drafts an internal update from a defined Jira scope and the corresponding Confluence decision pages. Its output is easy for a release owner to verify.

If an existing agent already fits, try that first. Open Chat's agent menu or the Studio agent library, inspect the agent's details, and start a conversation with it. You do not have to build a custom agent to use Rovo agents. How to browse available agents.

Use this five-day plan for a custom pilot:

1. Monday: choose the job and owner. Pick one release, its source pages and the person who approves the brief. Have the admin check Rovo access, the verified business-domain requirement, and your permission to create agents in Studio. Write down how long the existing preparation process takes.
2. Tuesday: create the agent. Open App switcher → Studio → Agents → Create under Rovo agent. Describe its role, settle its name, instructions and conversation starters, then select Next. Review knowledge, tools and reasoning settings. The guided flow finishes with Create; manual setup finishes with Activate. Agent creation instructions.
3. Wednesday: test a draft-only job. Supply the Jira scope and Confluence references. Start without write tools. Ask for completed work, unresolved blockers, decisions and missing evidence. Try a normal release, conflicting documents and a question the source material cannot answer.
4. Thursday: test access and review. Try the agent with intended users who have different access. Check every linked source. Treat the knowledge selection and prompt as guidance about its task; use account and content permissions to control actual access.
5. Friday: decide whether to keep it. Compare preparation plus review time against the baseline. Keep it if the owner can verify the brief and the combined work decreases. Add a single needed write action only after that test succeeds.

A starting instruction you can adapt:

Draft an internal release brief from the Jira scope and Confluence links I provide. Separate completed work, open blockers, decisions and unanswered questions. Link the source for each factual statement. Mark missing or conflicting evidence explicitly. Do not infer completion from an optimistic comment. Return a draft for review; do not publish or change work items.

That instruction is a proposed workflow, not a guarantee that the model will follow every rule. Removing unnecessary write tools and checking the output make it a useful pilot.

Choose the agent's account deliberately. With User's account, an interactive agent inherits the person using it. With Agent's account, it uses permissions assigned to its own identity. In automation, the user-account option uses the automation creator's permissions. Atlassian recommends the agent-account option for most automated jobs so access can be narrower. Agent identity and access rules.

A tool is an individual action, such as creating a page. A skill supplies reusable instructions for a multistep task. Neither a clever instruction nor an impressive agent name establishes what access it should have.

Once the pilot works, Studio can connect an agent to an event through Triggers → Add automation trigger, or through an automation flow's Use agent action. Current custom-agent flows can perform actions directly without confirmation at every step. Admin or user controls can block those writes; then later automation actions use the agent's returned text. Built-in agents also need subsequent actions in the documented flow examples. Current agent automation behavior.

Six Workflows Worth Trying, Ranked by Practical Value

The best candidates have a recurring input, a recognizable output and someone who can judge whether it is right. These are proposed uses, not measured savings.

My first choice is the release brief. It combines both apps, repeats often and produces an artifact whose owner already knows how to inspect. Organization-wide knowledge cleanup has a much less forgiving definition of success.

Admin and Data Controls to Check Before Expansion

The important controls govern different things: who can use AI, who can build, what an agent can access, where data is stored and how contributed data is used.

AI access is not a data exclusion list. In Atlassian Administration, use AI and Rovo → Access to manage app access. Blocking an app disables its AI experiences, but if AI remains active in another app on the site, Rovo can still use accessible data across the site. Source permissions are still essential. Rovo access controls.

Review Studio creation rights. Studio app admins can use Settings → Studio settings → Agents to allow all users, selected groups or only Studio app admins to create agents. The selected-group option supports up to ten groups. Studio creation permissions. In migrated organizations, also review the Studio roles carried over from Jira, Confluence and Jira Service Management: Studio app admins receive full creation rights. What changes with the standalone Studio app.

Separate provider training from Atlassian data contribution. Atlassian says its third-party model providers neither retain customer inputs and outputs nor use them to train their services. Default processing can nevertheless send data to those providers. Eligible Cloud Enterprise organizations can request Atlassian-hosted-only models. Non-AI Search also remains available when AI features are disabled. Atlassian's AI Trust Center.

Rovo itself stores Chat prompts and responses for 30 days in conversation history. Chat history and retention.

Atlassian has its own contribution policy. It can use de-identified, aggregated metadata, including common phrases and topics extracted from Rovo conversations, to improve experiences; contributed metadata can fine-tune open models inside Atlassian. The policy changed on August 17, 2026. Data contribution FAQ.

Check Administration → Security → Data contribution. The highest active plan in the organization, including trials, determines the defaults:

Specific compliance arrangements can have different defaults. Inspect the actual organization settings rather than assuming “no third-party training” answers the whole question. Contribution defaults and available controls.

Check residency scope separately from processing. Rovo supports regional storage for in-scope data, including chat session logs, agent configuration, bookmarks and ingested third-party content. Source code and configuration for apps built in Studio with Forge, Atlassian's app development platform, are outside that scope. A storage location is not a promise that every AI request is processed there. Rovo and Studio data residency scope.

Two Useful Things to Build Around Rovo

The opportunity is a reliable team workflow with maintained sources and evaluation, not another general-purpose assistant. These are implementation ideas for an internal team or a specialist serving Atlassian customers.

Release-brief implementation: the strongest option. An engineering operations lead could fund a reusable brief agent, an agreed output template, an owner review and a record of corrections. The smallest useful version covers one release process with draft-only output. DataForSEO's November 5, 2026 lookup estimates 1,300 U.S. monthly Google searches for “jira automation”. That is evidence of interest in the broader job, not proof that people will buy this specific agent. The catch: Rovo already supplies the building blocks. The value must come from fitting the customer's process and keeping the output dependable.

A maintained onboarding knowledge pack. An IT or enablement owner could fund a curated set of Confluence sources, a role-specific agent and a test set of questions new joiners actually ask. The smallest version covers one team and routes unanswered questions to a named knowledge owner. The same DataForSEO lookup estimates 2,900 U.S. monthly searches for “knowledge management”, a broad category signal. The catch: the work is largely document ownership and maintenance. An agent over contradictory pages merely gives those contradictions a conversational interface.

Common Rovo Questions

How do you open Rovo in Jira?

Select Ask Rovo near your profile avatar, or the Rovo button at the bottom right. Supported editing surfaces also offer /ai or /rovo. If the experience is unavailable, have your admin check access.

Is Atlassian Rovo an MCP?

Rovo is the wider AI experience. Atlassian MCP Server is a connection interface that lets external AI clients work with Atlassian data. MCP means Model Context Protocol. You do not need to configure that connection to use Rovo inside Jira or Confluence. Atlassian MCP Server overview.

Can Rovo work with Jira or Confluence Data Center?

There are Data Center connectors, but the full Rovo AI experience requires an eligible Cloud subscription. Connecting Data Center knowledge does not turn Rovo into a locally hosted service. Data Center access requirements.

What are useful Rovo agent examples?

A release-brief drafter, backlog-review helper and proposal reviewer are sensible starting ideas. Choose the one with accessible source material and an owner who can judge the result. A narrow recurring job is easier to evaluate than an assistant expected to answer anything.

Honest Limits and Your Monday Move

Rovo does not resolve contradictory documentation, decide which source your business should trust or make a generated statement authoritative. Treat its output as a draft, including when it sounds certain. An answer limited by one person's access can also miss information another teammate considers essential.

Actions have concrete limits. Atlassian currently documents 20 work items per bulk Jira action, and its Edit page tool appends content at the bottom of a Confluence page. Do not assume it can precisely rewrite any section. In interactive agent chat, consequential tools request confirmation; an automated flow has different execution behavior. Agent tools, confirmation and limits.

Included plans still have usage allowances. Monitor the organization's pooled Rovo credits and connector indexing capacity before expanding a successful pilot. How Rovo credits work. A connector's presence also does not prove that every file is available to answer a question.

Monday's move: the release owner and Atlassian admin choose one real handover, verify its sources and access, then build a draft-only Release Brief agent. Keep it only if checked output reduces the combined work. If nobody owns the source quality or can review the result, fix that first.

If you want a scoped Rovo workflow built around your team's process, get help with AI agent development.