atlassian/atlassian-mcp-server

Atlassian MCP Server

The official Model Context Protocol (MCP) server for Atlassian: a cloud-hosted bridge that gives your AI tools secure, real-time access to Jira, Confluence, Jira Service Management, Bitbucket, Compass, Loom, and the wider Atlassian platform — powered by your Teamwork Graph.

Getting started · Supported tools · Security & admin · Community

---

The official Atlassian MCP Server is a cloud-based bridge between your Atlassian Cloud site and compatible external tools. Once configured, it enables those tools to interact with Jira, Confluence, Jira Service Management, Bitbucket, Compass, Loom, and Atlassian platform data (Projects, Goals, Teams, Focus, Talent, and the Teamwork Graph) in real time. Authentication uses OAuth 2.1 or API tokens, so every action respects the user's existing access controls.

With the Atlassian MCP Server, you can:

• Summarize and search Jira, Jira Service Management, Confluence, Bitbucket, Projects, Goals, and more without switching tools.
• Retrieve and review recordings of videos and meetings from Loom.
• Create and update work items or pages based on natural language commands.
• Automate repetitive work, like generating work items from meeting notes or specs.

Connect once, then describe what you want — no tab switching and no copy-pasting. Your AI selects and runs the right actions.

It's built for developers, content creators, and project teams who work in IDEs or AI tools and want to use Atlassian data without constantly switching context.

[!IMPORTANT]

v2 is now the recommended version. New setups should use https://mcp.atlassian.com/v2/mcp, which exposes more tools and more products. Existing v1 connections automatically start to expose and use v2 tools; incompatible clients may need to clear cached client IDs or .well-known credentials to keep authenticating. For setup steps, see Getting started with the Atlassian MCP Server.

One-click setup

Pick your AI client below to install the official Atlassian MCP Server. Each button uses your client's native install link, so you don't need to edit any JSON config by hand.

Add to Cursor Reference issues and log work in your codebase.

Add to VS Code Search and create Jira issues via GitHub Copilot.

Add to ChatGPT Search, summarize, and create right from ChatGPT.

Add to Claude Bring Jira and Confluence into Claude workflows.

Let your agent do the setup

Most AI coding agents can install, authenticate, and configure the server themselves. Copy and paste this prompt into your agent:

Set up Atlassian MCP for this agent using the official setup guide at
https://support.atlassian.com/atlassian-rovo-mcp-server/docs/getting-started-with-the-atlassian-remote-mcp-server/
and the MCP server URL https://mcp.atlassian.com/v2/mcp. Then start the Atlassian MCP
authentication flow so I can sign in.

Or add the server manually with your client's own command:

| Client | Command or configuration | | --- | --- | | Claude Code | claude mcp add --transport http atlassian https://mcp.atlassian.com/v2/mcp, then run /mcp in a session to authenticate | | Codex | codex mcp add atlassian --url https://mcp.atlassian.com/v2/mcp | | Claude Desktop | Settings → Extensions → Browse extensions → Plugins, then search for Atlassian | | Codex Desktop | Plugins or Connectors, then install Atlassian Rovo | | VS Code / GitHub Copilot | Extensions view → search @mcp Atlassian → Install | | Cursor | Atlassian plugin for Cursor with MCP → Add to Cursor | | Windsurf | Cascade → MCP servers → Add Server → Add custom server, with serverUrl set to https://mcp.atlassian.com/v2/mcp | | Any other MCP client | Use the server URL https://mcp.atlassian.com/v2/mcp |

Contents

• One-click setup
• Supported clients
• Plugin packaging and compatibility
• Supported products and tools
• How the server exposes tools
• MCP gateways
• Before you start
• Data and security
• How it works
• Example workflows
• Tips and tricks
• Admin notes: managing access
• Security
• Support and feedback
• Disclaimer

---

Supported clients

The Atlassian MCP Server works with a growing list of MCP-compatible clients:

| Client | Setup reference | | --- | --- | | OpenAI ChatGPT | Connectors / MCP guide | | OpenAI Codex (CLI and Desktop) | Codex MCP docs | | Claude (Claude.ai, Desktop, and Code) | Claude MCP docs | | Cursor | Atlassian on the Cursor marketplace | | Visual Studio Code (GitHub Copilot) | VS Code MCP docs | | GitHub Copilot CLI | About Copilot CLI | | Google Gemini CLI | Gemini CLI MCP docs | | Windsurf | Windsurf MCP docs | | Docker | Docker MCP Catalog and Toolkit | | Kiro and other Agent Plugins clients | Kiro Powers documentation | | Amazon Quick Suite | MCP integration guide |

The Atlassian MCP Server also supports any local MCP-compatible client that can run on localhost and connect to the server via the mcp-remote proxy. This enables custom or third-party integrations that follow the MCP specification.

[!TIP]

For the current, canonical list of supported clients and step-by-step setup, see Getting started with the Atlassian MCP Server. You can also refer to your client's own MCP documentation or built-in assistant.

---

Plugin packaging and compatibility

This repository publishes the same MCP server and skills in several package formats so clients can use their native discovery and installation flows:

| Format | Manifest and configuration | Shared components | | --- | --- | --- | | Agent Plugins v1 | plugin.json and mcp.json | skills/ | | Claude Code plugin | .claude-plugin/plugin.json and .claude-plugin/marketplace.json | .mcp.json and skills/ | | Cursor plugin | .cursor-plugin/plugin.json | .mcp.json and skills/ | | Gemini extension | gemini-extension.json | MCP server configuration embedded in the manifest | | MCP Registry | server.json | Remote server metadata |

The skills in skills/ target the /v2/mcp endpoint and are what every packaged plugin installs. The previous generation is archived in skills/v1/ for anyone still on a v1 endpoint; it is not installed by any plugin and is set up manually. See skills/README.md for the differences.

Both mcp.json and .mcp.json are intentional. Agent Plugins requires the root mcp.json filename and its portable transport vocabulary; existing clients continue to use .mcp.json and their native configuration vocabulary. Keep their endpoint settings aligned when changing either file.

The root package can be imported by clients that implement Agent Plugins, including Kiro Powers. Client-specific manifests remain available and are not replaced by the portable package.

---

Supported products and tools

Tools are organized by product and intent (read, write, search, delete, or manage). Organization admins grant or revoke access at the permission-group level, and each tool inherits the access of its parent group.

| Product | Permission groups | OAuth 2.1 | API token | | --- | --- | :---: | :---: | | Jira | read_jira · write_jira · search_jira | ✅ | ✅ | | Jira (admin-enabled) | delete_jira · manage_jira | ✅ | ✅ | | Confluence | read_confluence · write_confluence · search_confluence | ✅ | ✅ | | Bitbucket Cloud | read_bitbucket · write_bitbucket | ✅ | ✅ | | Jira Service Management | read_jsm · write_jsm | — | ✅ (only) | | Atlassian platform | read_teamwork_graph · write_teamwork_graph · search_atlassian | ✅ | ✅ | | Loom | read_loom · write_loom | ✅ | ✅ | | Goals | read_goals · write_goals | ✅ | ✅ | | Projects | read_projects · write_projects | ✅ | ✅ | | Teams | read_teams · write_teams | ✅ | ✅ | | Focus | read_focus · write_focus | ✅ | ✅ | | Talent | read_talent · write_talent | ✅ | ✅ | | Compass | read_compass · write_compass | ✅ (only) | — |

Product-specific conditions:

• delete_jira and manage_jira are disabled by default and must be enabled by an admin before they can be used.
• Bitbucket Cloud tools are only available if your Bitbucket workspace is linked to an organization. Workspaces that aren't linked to an organization can't be selected when authenticating.
• Jira Service Management tools only support authentication via API token, and are only available if an organization admin has enabled API token authentication.
• Loom tools are only available for Loom workspaces linked to an Atlassian site.
• Teamwork Graph tools can retrieve data from third-party services connected to Jira, such as linked pull requests, builds, and deployments. For GitHub for Atlassian, full access means tools retrieve GitHub data based on the user's GitHub permissions, while limited access means they retrieve data based on the user's Jira permissions. Azure DevOps, GitLab, Jenkins, and Spinnaker connectors follow the limited access model.

[!NOTE]

For the complete, current tool reference, including the required scope for each permission group, see Supported tools.

---

How the server exposes tools

Rather than sending every tool definition to the AI client at connection time, the Atlassian MCP Server exposes a small set of the most-used tools and lets the client discover the rest on demand. Tools marked Primary are visible directly to agents.

This has two benefits:

• Your client's context stays free for the task at hand instead of being filled with tool definitions.
• New tools become available without the client needing to reconnect.

MCP gateways

If you're using an MCP gateway, or otherwise need every tool available up front rather than using the discovery and execute methods, use the tools=all override to expose all tools as a paginated flat tool list:

https://mcp.atlassian.com/v2/mcp?tools=all

---

Before you start

Check that your environment meets these requirements before you set up the server.

Prerequisites

The requirements depend on how you connect:

For supported clients

• An Atlassian Cloud site with one or more of Jira, Confluence, Jira Service Management, Bitbucket, or Compass
• Access to the client of choice
• A modern browser to complete the OAuth 2.1 authorization flow, or API token credentials for headless authentication

If your organization admin has disabled authentication via API token, MCP clients can't connect with a token and need to use OAuth 2.1 instead.

For IDEs or local clients (desktop setup)

• An Atlassian Cloud site with one or more supported products
• A supported IDE (for example, Claude Desktop, VS Code, or Cursor) or a custom MCP-compatible client
• Node.js v18+ installed to run the local MCP proxy (mcp-remote)
• A modern browser for completing OAuth login, or API token credentials for headless authentication

---

Data and security

The server enforces several security controls:

• All traffic is encrypted in transit over HTTPS (TLS 1.2 or later), per Atlassian's security practices.
• OAuth 2.1 and API token authentication provide secure access control.
• Data access respects user permissions across every connected Atlassian product, including Jira, Confluence, Jira Service Management, Bitbucket, Compass, and Loom.
• If your organization uses IP allowlisting for Atlassian Cloud products, tool calls made through the Atlassian MCP Server also honor those IP rules.

For a deeper overview of the security model and admin controls, see:

• Understand Atlassian MCP Server
• Control Atlassian MCP Server settings

---

How it works

Architecture and communication

1. A supported client connects to the server endpoint. The recommended endpoint for all clients is:

   https://mcp.atlassian.com/v2/mcp
   

The v1 endpoints (https://mcp.atlassian.com/v1/mcp/authv2 and https://mcp.atlassian.com/v1/mcp) remain supported, and existing v1 connections automatically start to expose and use v2 tools. 2. Depending on your setup, a secure browser-based OAuth 2.1 flow is triggered, or API token authentication is used. 3. Once authorized, the client streams contextual data and receives real-time responses from your connected Atlassian products.

[!WARNING]

After 30 June 2026, the legacy Server-Sent Events endpoint (https://mcp.atlassian.com/v1/sse) will no longer be supported. Update any custom clients configured to use /sse so they point to /mcp — preferably https://mcp.atlassian.com/v2/mcp.

[!NOTE]

Clients that can't complete authentication after the move to v2 tools may need to clear cached client IDs or .well-known credentials.

Permission management

Access is granted only to data that the user already has permission to view in Atlassian Cloud. All actions respect existing project or space-level roles. OAuth and API token authentication both honor configured scopes and Atlassian permissions.

API token authentication (headless)

API token authentication is available for headless, service-style, or non-interactive client setups (for example, backend systems or automations). It is also required for Jira Service Management tools.

Two mechanisms are supported:

| Mechanism | Header | | --- | --- | | Personal API token (Basic auth) | Authorization: Basic | | Service account API key (Bearer token) | Authorization: Bearer <api_key> |

• Admin enablement required: An organization admin must enable API token authentication for the Atlassian MCP Server (Atlassian Administration → Rovo → Rovo MCP server → Authentication).
• Scoped token required: Create a personal API token, or ask your admin for a service account API key, with the scopes required for the tools and data you need to access.
• Configuration guide: Configure authentication via API token
• Admin setting reference: [Control Atlassian MCP Server settings: Configure authentication](https://support.atlassian.c