Rovo is the name Atlassian put on its whole AI push into Jira and Confluence — and the name is also the problem. Atlassian ships at least four different things under the “Rovo” brand — the AI layer itself, a separate coding agent, a protocol server, and a custom-agent builder — and the marketing rarely draws the line between them clearly. If you’ve landed here trying to figure out what Rovo actually is before deciding whether it matters for your team, this page draws that line, then points you to the dedicated page for whichever piece you actually need.

What is Rovo?

Rovo is Atlassian’s AI layer across Jira and Confluence: search, chat, and agents, built on what Atlassian calls the Teamwork Graph — a structure that links people, projects, work items, and content so the AI has real organizational context instead of guessing from a blank prompt (atlassian.com/software/rovo). It ships in three pieces, per Atlassian’s own “What is Rovo” doc:

  • Rovo Search — one query across Atlassian apps and connected third-party tools, respecting existing permissions, so you’re not context-switching between five search boxes.
  • Rovo Chat — a conversational interface, available inside Jira, Confluence, and a browser extension, for asking questions and getting drafts grounded in company data you already have access to.
  • Rovo Agents — AI teammates that generate content, make recommendations, and automate tasks like cleaning up a Jira backlog or organizing Confluence pages.

Critically, Rovo is not a separate app you go buy — it’s a capability bundled into your existing Jira/Confluence subscription. “Rovo credits are included in all paid Jira, Confluence, Service Collection and Teamwork Collection cloud subscriptions,” per Atlassian’s docs, with the allowance scaling by plan tier: 25 credits per user per month on Standard, 70 on Premium, 150 on Enterprise. Full mechanics — what a credit actually costs to spend, and what happens when you run out — are covered in Rovo pricing and credits explained; the short version is below.

Rovo, Rovo Dev, Rovo MCP Server, Rovo Agents — what’s the difference?

This is the actual job of this page, because it’s where most confusion lives. Four names, four different things:

  • Rovo (above) — the AI layer inside Jira/Confluence itself: search, chat, agents. Metered by the credit allowance bundled with your plan.
  • Rovo Dev — a separate, standalone coding agent — a CLI, a VS Code extension, and PR code review — that reads and writes actual code, priced on its own at $20 per developer per month with its own credit pool, entirely apart from standard Rovo’s allowance. It’s the closest thing Atlassian has to Claude Code or Cursor’s agent mode, with native Jira/Confluence context as the differentiator. Full depth: Rovo Dev, explained.
  • Rovo MCP Server — the protocol-level way to expose the Teamwork Graph (Jira, Confluence, connected data) to external AI tools over the Model Context Protocol, so an agent like Claude Code or Cursor can read your Atlassian context without going through Rovo’s own chat UI at all. It’s connector infrastructure, not a chat product. Full guide: Rovo MCP Server, explained.
  • Rovo Agents — the automation capability within standard Rovo: pre-built or custom AI teammates you can build to run specific, repeatable jobs (backlog grooming, page cleanup, triage). It’s a feature of Rovo, not a separate product, but it’s easy to conflate with “an agent” in the generic AI sense.

There’s a fifth thing worth naming here too, because it comes up in the same searches: Atlassian’s DevOps connectors (GitHub, GitLab, Azure DevOps) that feed source-control context into the Teamwork Graph, plus Rovo for GitHub Copilot, a complement rather than a competitor to Copilot. That whole map — connectors, Bitbucket, and the “three Copilots” confusion — is in Rovo DevOps integration, explained.

If you’re comparing Rovo specifically against Microsoft’s tools rather than untangling Atlassian’s own naming, see Rovo vs Copilot for Jira teams — they solve different problems (Rovo acts inside Jira; Microsoft 365 Copilot only reads it via a connector) and aren’t really substitutes for each other.

If your next question is “okay, but is it safe to turn on,” see Rovo security: the pre-rollout admin checklist — the concrete decisions (connector blocklists, data residency, disconnect behavior, audit scope, permission inheritance) behind the vague “yes, SOC 2” answer most partner blogs stop at.

What does Rovo cost?

Kept brief here as orienting context — the real depth, including what actually burns a credit and what happens at the limit, lives on the dedicated pricing page. The shape of it: Rovo has no separate price tag for most teams. It’s bundled into paid Jira and Confluence Cloud plans, and instead of a per-seat AI fee you get a pooled monthly credit allowance — 25 credits per user on Standard, 70 on Premium, 150 on Enterprise — that meters agent runs and chat usage across your whole org. Rovo Dev is the one exception, billed entirely separately at $20/developer/month. If you want to size what a real team would actually draw down before committing to a plan tier, run it through the Rovo credit cost estimator.

Is Rovo worth it?

For most teams already paying for Jira or Confluence Cloud, yes, in the sense that there’s very little to lose — the credit allowance is bundled in, so turning on Rovo Search and Chat costs nothing extra to try, and the Teamwork Graph context makes answers noticeably less generic than a bare LLM prompt with no access to your actual tickets and docs. Where it gets genuinely worth configuring, rather than just leaving switched on, is Rovo Agents: the value is concentrated in a handful of repetitive, well-defined jobs (backlog grooming, intake triage, page cleanup), not in treating Rovo like a general-purpose chatbot. And if your team’s real ambition is agents that act outside Jira/Confluence too — reading a Jira ticket and also touching your codebase, your CRM, or a customer database — that’s the point where standard Rovo’s scope runs out and you’re either looking at Rovo Dev (for code) or a custom-built agent via the MCP layer.

If the agent you eventually want doesn’t exist yet as a standard Rovo capability, we build them.