Atlassian Rovo
Also known as: Atlassian, Atlassian Corporation, TEAM, Rovo, Rovo Agents, Rovo Studio, Rovo Chat, Rovo Search, Rovo Dev, Rovo Desktop, Atlassian Rovo MCP Server, Teamwork Graph, Jira Service Management, JSM, Rovo Ops, Rovo Service, Agents in Jira, Teamwork Collection, Service Collection, Atlassian Intelligence
Atlassian Rovo is the agent layer across Jira, Confluence, Jira Service Management and Loom. Teams assign Jira work items to agents, start them from automation rules and workflow transitions, and build their own in Rovo Studio from subagents, skills, tools and knowledge in the Teamwork Graph. It ships 26 ready agents, scores agents against test prompts with an LLM judge, and pins chat logs and agent settings to a chosen data residency location. Rovo comes with paid Atlassian Cloud plans and is metered in credits.
Atlassian makes Jira, Confluence, Jira Service Management, Loom and Bitbucket, and Rovo is the AI layer that runs through all of them, with search, chat and agents grounded in the same work data. It now spans Rovo Studio, where teams build agents, automations and apps in plain language, agents that take Jira work items like any teammate, Rovo Dev for coding, and Rovo apps for the desktop, mobile, Chrome and Edge, Slack and Microsoft Teams. Atlassian reports that 90 percent of its enterprise cloud customers use Rovo, and Rovo Standard, in beta, offers it to people without a paid Atlassian app.
An agent is built through chat or by hand from instructions, subagents that each carry their own knowledge from Confluence spaces, Jira projects or Google Drive folders, skills that orchestrate tools, and a reasoning tier, where Think deeper agents can pick from models Atlassian has tested. Work reaches agents when someone assigns them a work item, mentions them in a comment or moves work through a workflow transition, and automation rules call them on a schedule, a published page or a new work item. Answers draw on the Teamwork Graph, which indexes Atlassian and connected third party content under each user's permissions.
Skills create and edit work items and pages and send Gmail, Outlook and Teams messages once the person who prompted them confirms. Rovo ships 26 agents, from a Release Notes Drafter and Jira Triage Agent to the Ops Expert and Request Router, plus nine third party MCP agents from Figma, Canva, HubSpot and others. Owners score agents against datasets of up to 100 prompts with an LLM judge, debug the subagent, tools and model behind each answer and read live transcripts, and the Rovo MCP Server opens more than 215 tools to Claude, ChatGPT, Cursor and VS Code.
Rovo is included in the Standard, Premium and Enterprise Cloud plans of Jira, Confluence, Loom and Jira Service Management and in the Teamwork and Service Collections, where Teamwork Collection Standard is $14.02 a user a month and Premium $28.76. Each plan adds a pooled monthly allowance of 25, 70 or 150 credits a user on single apps and 250, 700 or 1,500 on the collections. Basic chat and agent events cost 10 credits, Think deeper and advanced agents vary, and extra usage bills at $0.01 a credit from December 3, 2026, unless an admin caps it or turns it off.
Rovo holds SOC 2 and ISO 27001 certifications, supports HIPAA compliant use and keeps chat and agent inputs for 30 days, and it runs on OpenAI, Anthropic and Google models alongside Atlassian hosted open source ones, whose providers neither retain nor train on customer data. Data residency pins chat logs, agent configuration and ingested third party content to locations from the EU and Germany to Japan and the US. Rovo runs only on Atlassian Cloud and is unavailable on Atlassian Government Cloud.
Rovo fits organizations already running Jira, Confluence or Jira Service Management in the cloud that want agents working inside the tracker and wiki their teams use, service desks that want triage and request routing on the same platform, and admins who want agent creation, evaluation and verification governed in one place. Its agents act through product APIs and connectors, so work that needs a browser or desktop app driven click by click sits outside them, and Data Center estates reach Rovo only through connectors on a Cloud plan. Think deeper and advanced agents draw more credits, so autonomous agents on busy automation rules need a spending cap.
Vendor details
Canonical URL
https://www.atlassian.com/software/rovo
Category
Enterprise operations agent
Subcategory
Agents, chat and search across Jira, Confluence, Loom and connected apps, built in Rovo Studio and grounded in the Teamwork Graph
Funding status
Product of Atlassian (NASDAQ: TEAM).
Company status
first party product
Use cases & customers
Primary use cases
Target customers
Deployment options
Integrations
Rovo works across Jira, Confluence, Jira Service Management, Loom and Bitbucket and indexes connected apps such as Google Drive, SharePoint and GitHub into the Teamwork Graph under each user's permissions. Skills send Gmail, Outlook, Microsoft Teams and Slack messages and create Google Docs and Calendar events, and nine third party MCP agents from Figma, Canva, Box, Amplitude, HubSpot, Intercom, Gamma, Replit and Lovable run inside Jira and Confluence. The Rovo MCP Server opens more than 215 tools to Claude, ChatGPT, Cursor, VS Code and GitHub Copilot.
In practice
A service team adds the Request Router to its Jira Service Management project as an automation agent. Each new request is triaged and its fields updated before a person picks it up, and the Service Request Helper drafts the reply.
A product lead assigns a batch of bug work items to the Jira Coding Agent, which turns each one into code in a cloud sandbox and opens a pull request for engineers to review.
Before changing a support agent's instructions, its owner uploads a CSV of 80 past questions with expected answers, runs a response accuracy evaluation and compares the LLM judge's scores with the previous version before publishing.
Sources & related URLs
Related / legacy domains
Research sources
Agentic Index coverage score
13.5 / 14 capabilities · 96%
| Integrations & Tool Calling | Full |
|---|---|
|
Rovo's skills create, edit, assign, comment on and transition Jira work items, publish and add to Confluence pages, create Google Docs and Calendar events, send Gmail, Outlook, Microsoft Teams and Slack messages, and comment on Bitbucket pull requests, all within the user's permissions. Agents take custom tools and reach data in connected third party apps, whose permissions Rovo keeps in sync. Nine out of the box MCP agents from Figma, Canva, Box, Amplitude, HubSpot, Intercom, Gamma, Replit and Lovable call each vendor's own MCP server once an admin enables it on Premium or Enterprise. Sourcesupport.atlassian.com/rovo/docs/chat-actionsread 2026-10-24 |
|
| Workflow Orchestration | Full |
|
An agent is built through chat or by hand from instructions, subagents that each handle a different situation with their own instructions and knowledge, skills that orchestrate tools with reasoning, and up to five recommended tools. Quick answers allows 10 reasoning loops and Think deeper 60 for analysis and synthesis. Agents run as steps inside automation flows, where later steps use the agent's response, and Max mode in Rovo Desktop takes a complex task through its steps, reads errors, adjusts its approach and delivers an output. Sourcesupport.atlassian.com/rovo/docs/create-and-edit-agentsread 2026-10-24 |
|
| Knowledge Grounding & RAG | Full |
|
Each subagent carries its own knowledge sources, such as Confluence spaces, Jira projects or a Google Drive folder, and answers draw on the Teamwork Graph, an index of Atlassian and connected third party content that follows each app's permissions and updates when they change. Admins can block parts of Google Drive and SharePoint from indexing, and disconnecting an app deletes its indexed content within 30 days. Deep research mode takes larger questions across sources, and web search adds public material. Sourcesupport.atlassian.com/rovo/docs/rovo-data-privacy-and-usage-guidelinesread 2026-10-24 |
|
| Human Oversight & Guardrails | Full |
|
Skills act only after the person who prompted Rovo confirms, deleting a work item asks for a second confirmation, and every action respects that person's permissions. In automation rules agents can run without per action confirmation, and admins or users can block agents from acting there, so write tools fail and later steps work from the agent's response. Max mode in Rovo Desktop asks before sensitive local actions unless the user turns on Auto approve, which Atlassian warns to use with extreme caution, and admins can verify agents for the whole organization. Sourcesupport.atlassian.com/rovo/docs/chat-actionsread 2026-10-24 |
|
| Security, Identity & Governance | Full |
|
Rovo and Atlassian's AI features hold SOC 2 and ISO 27001 certifications, Rovo can be used in a HIPAA compliant way, and chat and agent inputs and outputs are kept for 30 days for safety and security while model providers neither retain them nor train on them. Agents return only what the user can access, Studio admins decide who may create agents, owners add editors and managers, and an agent can be limited to named people. For the Rovo MCP Server, admins set what can be read, written or deleted, block AI access by site, space or classification, and route sign in through their own identity provider. Sourcesupport.atlassian.com/rovo/docs/rovo-data-privacy-and-usage-guidelinesread 2026-10-24 |
|
| Observability & Auditability | Full |
|
Anyone chatting with an agent can open a debug view showing the subagent it picked, the tools it used, their inputs and outputs, and the request and model IDs, and automation flows carry an audit log with a debug log for the person who built them. Agent owners and managers read live conversation transcripts with thumbs up or down feedback and categories such as Inaccurate, kept for 30 days after the last message, and track how often each agent is used. Rovo Desktop writes tool activity to local app logs. Sourcesupport.atlassian.com/rovo/docs/debug-responses-from-rovo-agentsread 2026-10-24 |
|
| Memory & State Persistence | Full |
|
Rovo keeps two kinds of memory for each person. An implicit summary built from Teamwork Graph activity across Atlassian apps can be corrected through suggested edits that apply within 24 hours, and explicit memories saved from chat instructions can be viewed and deleted one by one, after which Rovo behaves as if the fact never existed. Every memory is scoped to the user and to one Atlassian site, and organization admins cannot see what Rovo has stored. Rovo Dev CLI keeps a separate memory for coding sessions. Sourcesupport.atlassian.com/rovo/docs/what-is-rovo-memory-managementread 2026-10-24 |
|
| Deployment & Data Residency | Full |
|
Organization admins can pin Rovo's chat session logs, agent configuration, bookmarks and ingested third party content to a location through the same self serve migration as other Atlassian Cloud apps, choosing among Australia, Canada, the EU, Germany, India, Japan, Singapore, South Korea, Switzerland, the United Kingdom, the USA or Global. Indexed Atlassian content stays with its source app's location, while source code of Forge apps built in Studio sits outside residency. Rovo runs only on Atlassian Cloud Standard, Premium and Enterprise and is unavailable on Atlassian Government Cloud. Sourcesupport.atlassian.com/security-and-access-policies/docs/understand-data-residencyread 2026-10-24 |
|
| Prebuilt Agents, Templates & Packs | Full |
|
Rovo ships 26 ready agents, among them a Product Requirements Guide, Release Notes Drafter, OKR Generator, a Jira Coding Agent that turns work items into code in a cloud sandbox, Jira Triage Agent, Ops Expert, Service Request Helper and Request Router. Nine third party MCP agents come ready to enable, Marketplace partners add installable agents, and Rovo Desktop runs prebuilt tasks such as action items from a Loom video. Any agent can be duplicated and adapted with new instructions, tools and knowledge. Sourcesupport.atlassian.com/rovo/docs/atlassian-agentsread 2026-10-24 |
|
| Triggers & Channel Coverage | Full |
|
Automation triggers start agents when a page is published, a work item is created or a schedule comes round, such as every 2 days at 8:00am, and an agent can sit on a workflow transition or a board column so a status change brings it in. People assign a work item to an agent or mention it in a comment, call it with /rovo or /ai in the editors, and chat with agents in Rovo Chat, Studio, Loom, Slack, Microsoft Teams, the browser extension on any public page, and the desktop and mobile apps. Sourcesupport.atlassian.com/rovo/docs/agents-in-automationsread 2026-10-24 |
|
| Model Flexibility & Routing | Full |
|
Rovo draws on Atlassian hosted open source models and hosted models from OpenAI, Anthropic and Google. Each reasoning tier has a default model Atlassian has tested, Think deeper agents can choose from a list Atlassian reassesses for cost and performance, and a deprecated model brings at least 30 days' notice before an agent moves to the nearest equivalent. Advanced agents with model selection bill at variable credit rates, and in Rovo Dev CLI the /models command switches models, each with a credit multiplier. Customers cannot bring their own model or provider account. Sourcesupport.atlassian.com/rovo/docs/set-a-reasoning-tier-for-a-rovo-agentread 2026-10-24 |
|
| APIs, SDKs & MCP Extensibility | Full |
|
The Rovo MCP Server opens more than 215 tools across 10 or more Atlassian apps to clients such as Claude, ChatGPT, Cursor, VS Code and GitHub Copilot, and a Teamwork Graph CLI, in open beta, gives coding agents such as Claude Code and Codex the same context under admin scopes. Rovo Dev CLI connects to MCP servers, runs in a server mode other tooling can drive and takes Agent Skills. Searches through the CLI or the MCP server draw credits, while installing either costs nothing. Sourceatlassian.com/platform/rovo-mcpread 2026-10-24 |
|
| Testing, Debugging & Optimization | Full |
|
Agent owners build datasets of up to 100 prompts, typed in or uploaded as a CSV with an optional column of expected responses, and run up to three evaluations at once. Response accuracy compares answers with the expected ones and resolution rate measures how often support requests get resolved, each scored by an LLM judge whose reasoning shows for every prompt, and a manual mode generates responses in bulk for people to review. Results can be compared after each change to an agent, and admins mark agents as verified for the organization. Sourcesupport.atlassian.com/rovo/docs/evaluate-the-performance-of-your-rovo-agentread 2026-10-24 |
|
| Browser & Computer Use | Partial |
|
Max mode in Rovo Desktop, a beta for Windows and macOS that runs on Rovo Dev credits, edits files and runs code and commands on the user's machine, and the desktop app can attach a screenshot of a chosen window to a prompt. The Chrome and Edge extension reads the page open in the current tab to answer questions, and the Jira Coding Agent writes code in a cloud sandbox. None of these click, type or navigate through a browser or desktop app on the user's behalf. Sourcesupport.atlassian.com/rovo/docs/use-max-mode-in-rovo-desktopread 2026-10-24 |
|
The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded
Product updates
Rovo Dev code review is renamed Code Reviewer and is now generally available with any paid Jira or Teamwork Collection license, with no separate Rovo Dev subscription. The agent reviews pull requests in Bitbucket Cloud and GitHub, leaves inline comments with suggested fixes, checks changes against acceptance criteria in linked Jira work items, and runs each review in a temporary isolated environment that is deleted afterward. Reviews draw on the organization's shared Rovo credit pool, with credit billing starting December 3, 2026.
Bears on: Agent capability
View sourceOrganization admins can now let service accounts connect to the Atlassian MCP server using OAuth 2.0 client credentials. A service account gets product agent-interface scopes, exchanges its client ID and secret for a 60 minute access token, and calls https://mcp.atlassian.com/v2/mcp as a Bearer token with no browser sign in. The Atlassian Guard external sites policy now also applies when managed users connect through the Atlassian MCP server.
Bears on: Security / enterprise
View sourceSix tools on Atlassian MCP V2 now return rich UI cards in MCP clients instead of plain text: getJiraIssue, editJiraIssue, createJiraIssue, transitionJiraIssue and searchJiraIssuesUsingJql in Jira, and searchConfluence in Confluence. The rollout reaches customers in steps, and Atlassian asks anyone who does not see the cards by October 23 to contact support.
Bears on: Human approval / guardrails
View sourcePricing
Included in Standard, Premium and Enterprise Cloud plans of Jira, Confluence, Loom and Jira Service Management and in the Teamwork and Service Collections. Teamwork Collection Standard is $14.02 a user a month.
Per seat on the underlying Atlassian Cloud plan, with Rovo credits drawn per billable event from a pooled monthly allowance and extra usage billed per credit.
Included quota
Each paid plan adds a pooled monthly allowance per user: 25, 70 and 150 credits on Standard, Premium and Enterprise plans of Jira, Confluence and Loom, and 250, 700 and 1,500 on the Teamwork and Service Collections. Allowances from every eligible subscription combine into one pool, reset on the usage billing date and do not roll over. Rovo Search in Atlassian apps, definitions, summaries and write operations cost nothing.
What is public
Plan prices for the Teamwork Collection, the monthly credit allowance for each plan and collection, which events are free, basic or premium, the per credit list price and the alert and cap mechanics.
Billing mechanics
A per seat Atlassian Cloud subscription on Standard, Premium or Enterprise, with a Rovo credit allowance bundled into each tier and pooled across the organization. Service Collection is licensed per agent and Teamwork Collection per user. Extra usage beyond the allowance bills per credit, and admins can cap it or switch it off.
Cost watchouts
Service Collection is priced per licensed operator, not per employee raising tickets, so the AI budget follows how many operators are licensed. Premium interactions bill at variable rates that depend on reasoning steps, searches, tools and model power, and allowances on single app plans are a tenth of the collections' at each tier. Rovo needs a Standard, Premium or Enterprise Cloud plan, so a Data Center estate reaches it only through connectors on top of a Cloud subscription.
Variable cost rationale
Seats are a fixed per user fee on the underlying plan, and Rovo spends credits from a pooled allowance at 10 credits for basic chat and agent events and at variable rates for Think deeper, advanced agents with model selection and the Jira Coding Agent. Extra usage is on by default at $0.01 a credit, so autonomous agents on busy automation rules can run past the allowance until an admin cap stops them.
Additional watchouts
Extra usage is on by default, so a spending cap needs setting before autonomous agents go live on high volume automations.
Overage / add-ons
Extra usage is on by default at a list price of $0.01 per credit ($10 per 1,000), with billing for it taking effect December 3, 2026, and prepaid packs carry annual volume discounts. Admins can set a spending cap or switch extra usage off, after which billable interactions pause until the allowance resets. Org and billing admins get alerts at 80 and 100 percent of the allowance.
Sales call required
No, self serve available
Free / trial
Free trials of the Standard, Premium and Enterprise plans of Jira, Confluence and Jira Service Management. Rovo Standard is free during its beta for people without a paid Atlassian app, who move to Rovo Free afterward unless they opt in to the paid plan.
Lowest paid plan
Teamwork Collection Standard at $14.02 per user per month
Commercial notes
Atlassian bundles agent use into plans customers already buy and lists which events are free, basic at 10 credits or premium at variable rates. Prepaid credit packs carry annual volume discounts, and Rovo Standard brings Rovo to people outside paid Atlassian apps during its beta.
Key ambiguities
Rovo carries no price of its own, so its cost is set by the seat price of the plan or collection it comes with. Premium interactions are priced variably, Rovo Dev runs on a separate credit system, and the paid Rovo Standard price after its beta is not public.
Missing data
Exact credit costs for premium interactions, Rovo Dev credit pricing, Jira and Confluence seat prices, and the paid Rovo Standard price.
Related vendors
- 11th Estate — Agentic platform whose AI engine scans markets, matches an…
- Abacum — FP&A platform for growing companies whose AI Modeler builds, edits…
- Adopt AI — Finance agents that categorize, reconcile and close the books inside…
- Aera Technology — Decision intelligence platform where an always on agent executes…
- Akro AI — On premise operational intelligence that automates document heavy…
- alfred_ — AI executive assistant that triages the inbox overnight and scores…
Alternatives to Atlassian Rovo
The closest documented capability profiles to Atlassian Rovo among enterprise operations agents tracked by Agentic Index, ordered by similarity on the same 14 point evidence the rankings use. No vendor pays for placement.
- Microsoft14.0 / 14Fuller documented coverage on Browser & Computer Use
- Salesforce13.0 / 14A lighter documented profile than Atlassian Rovo
- Appian13.5 / 14Fuller documented coverage on Browser & Computer Use
- Boomi12.5 / 14A lighter documented profile than Atlassian Rovo
- Dust12.5 / 14A lighter documented profile than Atlassian Rovo
- Harvey12.5 / 14A lighter documented profile than Atlassian Rovo
Similarity is computed from each vendor's Agentic Index coverage score evidence, axis by axis, not from the totals. How this evidence is graded