Atlassian Rovo MCP Server

The Atlassian Rovo MCP Server is the official Model Context Protocol connection to Jira, Confluence, Jira Service Management, Bitbucket, Compass, and Loom, operated by Atlassian itself. Instead of switching between browser tabs, the ticketing system, and an AI assistant, a coding or research agent connects directly to your own Atlassian Cloud workspace and can read, search, and, depending on granted permissions, write there too. For teams already working in Jira and Confluence, this is a direct efficiency win: an agent can turn a meeting note into matching Jira issues, summarize the current state of a project from Confluence, or triage open Bitbucket pull requests, without anyone having to gather the information by hand.

What the server does

Per Atlassian, the server lets you search and summarize Jira issues, Confluence pages, Jira Service Management requests, Bitbucket repositories, Compass components, and Loom recordings, create and update issues or pages using natural language, and automate repetitive work such as generating work items from specs or meeting notes. The tool catalog is organized by product and permission group (read, write, search, delete, manage), so organization admins can precisely control which actions a connected AI client is allowed to perform.

Version, endpoint, and supported clients

Atlassian currently recommends version 2 at https://mcp.atlassian.com/v2/mcp, which covers more tools and more products than the earlier v1 version. Per the provider, existing v1 connections automatically start using v2 tools. The legacy SSE endpoint https://mcp.atlassian.com/v1/sse will no longer be supported after 30 June 2026. Supported clients include Claude (Claude.ai, Claude Desktop, Claude Code), OpenAI ChatGPT and Codex, Cursor, Visual Studio Code with GitHub Copilot, GitHub Copilot CLI, Google Gemini CLI, Windsurf, and any other MCP-capable client via the mcp-remote proxy. Many clients have ready-made one-click install links directly in the official repository.

Authentication and access rights

Sign-in uses OAuth 2.1 or API tokens; in both cases, only the permissions the signed-in user already has in Atlassian Cloud apply. API token authentication is particularly suited to non-interactive, headless automations and is required for Jira Service Management tools; it must be enabled in advance by an organization admin. Per Atlassian, all traffic is encrypted over HTTPS using TLS 1.2 or later, and any existing IP allowlists for Atlassian Cloud products also apply to calls made through the MCP server.

License and cost

The repository itself is licensed under Apache-2.0. The hosted MCP service does not add separate fees; it simply requires an existing Atlassian Cloud product (Jira, Confluence, etc.) with a valid plan, since the server only acts as a bridge to data you already have.

Who benefits most

The server is especially useful for development teams, product managers, and anyone who regularly switches between code, tickets, and documentation: IDE-based work with direct access to Jira context, automated issue creation from specs, or a quick summary of Confluence knowledge right inside a chat with an AI assistant. Someone who only checks Jira occasionally probably won't notice much difference — but anyone who works with Atlassian tools daily and relies on AI agents will save a lot of manual context switching.

An Atlassian Cloud product (Jira, Confluence, Jira Service Management, Bitbucket, Compass, or Loom) plus an MCP client that supports remote servers over Streamable HTTP (e.g. Claude Code, Cursor, Codex, VS Code with GitHub Copilot). Clients without native remote support can connect via the mcp-remote npm package.

For Claude Code, a single command is enough:

claude mcp add --transport http atlassian https://mcp.atlassian.com/v2/mcp

Then run /mcp in a session to start authentication. For Codex:

codex mcp add atlassian --url https://mcp.atlassian.com/v2/mcp

For Cursor, VS Code, ChatGPT, and other clients, the official repository provides ready-made one-click install links; alternatively, any MCP-capable client can be configured manually with the server URL https://mcp.atlassian.com/v2/mcp.

claude mcp add --transport http atlassian https://mcp.atlassian.com/v2/mcp

OAuth 2.1 (browser-based login for interactive clients) or API tokens (basic auth with a personal API token, or bearer auth with a service account API key, for headless/automated use). API token access must be enabled in advance by an organization admin; Jira Service Management supports API token auth only.

The server acts strictly with the permissions of the signed-in Atlassian account or API token. Access is granted per product through permission groups (read, write, search, delete, manage), which organization admins control centrally; delete and manage in Jira require separate enablement.

Requests run through Atlassian-hosted infrastructure at mcp.atlassian.com. Responses contain content from your own Atlassian workspace (issue text, Confluence pages, comments, Bitbucket metadata) and are returned to the MCP client and its language model. Per the provider, existing IP allowlists for Atlassian Cloud products also apply to calls made through the server.

With write-capable permission groups, an agent can create or edit issues and post comments; with delete rights, it can also remove content. Organization admins should scope permission groups as tightly as possible and enable delete/manage only where clearly needed. Treat API tokens like any other credential and keep them out of shared config files.