MCP Atlassian
Overview
MCP Atlassian is an MCP (Model Context Protocol) server implementation that connects AI assistants and IDE integrations to Atlassian products—specifically Confluence and Jira. It provides a standardized set of tools (e.g., search Confluence pages, get/create/update Jira issues) so LLM-based assistants can query, summarize, create, or modify content in Atlassian instances programmatically.
Key Features
- Supports both Atlassian Cloud and Server/Data Center (Confluence >= 6.0, Jira >= 8.14).
- Exposes read and write operations for Confluence and Jira through MCP tools (search, get page, create/update page, search issues, create/update issues, transitions, comments, attachments, etc.).
- Multiple authentication methods: Cloud API tokens, Server personal access tokens, and OAuth 2.0 (including a setup wizard and BYOT—Bring Your Own Token—mode).
- Distributed as a Docker image (ghcr.io/sooperset/mcp-atlassian:latest), suitable for running locally, in CI, or as a service for IDE integrations and AI assistants.
- HTTP transports available (SSE, streamable-http) for persistent connections and multi-user scenarios.
Authentication & Multi-tenant Support
MCP Atlassian supports:
- API Token authentication for Atlassian Cloud (recommended for simple setups).
- Personal Access Tokens for Server/Data Center.
- OAuth 2.0 (3LO) for Cloud with a built-in setup wizard to obtain refreshable tokens.
- BYOT: accept externally managed OAuth access tokens (server will not refresh BYOT tokens).
- Multi-cloud mode where each request can supply an Authorization bearer token and X-Atlassian-Cloud-Id allowing per-request tenant isolation.
Deployment & Integration
- Primary distribution is via Docker image; examples provided for running with environment variables or env-file.
- Can be run as an MCP service over stdio for IDE integrations or as an HTTP service (SSE or streamable-http) for multi-user access.
- Includes environment variables to filter spaces/projects, enable/disable tools, control logging, proxy settings, custom headers, and read-only mode.
- Designed for integration with IDEs and assistant clients (examples for Claude Desktop, Cursor, and MCP client libraries are provided).
Tools & Capabilities
MCP Atlassian exposes a clear tool set for AI assistants, including but not limited to:
- Jira: jira_get_issue, jira_search, jira_create_issue, jira_update_issue, jira_transition_issue, jira_add_comment, jira_download_attachments, agile/board/sprint helpers, etc.
- Confluence: confluence_search, confluence_get_page, confluence_create_page, confluence_update_page, get labels/comments, etc. Tool filtering and read/write controls are supported to limit surface area.
Use Cases
- Automatically create or update Jira issues from meeting notes or chatbots.
- Summarize and search Confluence documentation using LLMs, then create or update pages.
- Build chatops flows where an assistant finds relevant docs and files or performs triage and transitions on issues.
- Multi-tenant assistants where each user’s Atlassian token is used per request.
Security & Operational Notes
- Tokens and secrets should be kept out of repos and .env files should be protected.
- Custom HTTP headers are supported and masked in logs; SSL verification toggles exist for self-signed Server/DC deployments.
- When using BYOT or multi-tenant OAuth, responsibility for token lifecycle (refresh/rotate) may fall on external systems.
Contributing & License
- Project is MIT-licensed and welcomes contributions; repo contains CONTRIBUTING.md and SECURITY.md with guidelines and pre-commit hooks.
Summary
MCP Atlassian is a practical bridge between LLM-driven assistants and Atlassian tools, focusing on secure, configurable access to Confluence and Jira for both Cloud and Server deployments, and designed for easy integration via Docker and MCP transports.
