mcp-atlassian

Exposes 45+ Jira tools that map to the Jira REST API v3, including issue creation, retrieval, updates, and deletion with automatic field schema discovery. Uses a JiraClient mixin-based architecture that adapts payloads between Cloud (*.atlassian.net) and Server/Data Center deployments, handling custom fields, issue types, and project-specific field constraints through dynamic schema introspection rather than static field mappings.

Provides search operations that execute Jira Query Language (JQL) queries through the Jira Search API, returning paginated issue results with support for field projection, sorting, and result aggregation. Implements server-side filtering and result ordering to reduce payload size and network overhead, with built-in pagination handling for large result sets (>50 issues) that abstracts the complexity of offset/limit management from the caller.

Provides tools for creating, updating, and querying comments on Jira issues and Confluence pages with support for user mentions (@username) and automatic notification triggering. Uses the Jira/Confluence REST APIs to handle comment creation with mention parsing, automatic @-notification of mentioned users, and comment visibility settings (private, team, public). Comment queries return full comment history with author metadata, timestamps, and edit history, enabling AI agents to participate in issue discussions and track conversation context.

Provides tools for uploading files to Jira issues and Confluence pages, with automatic content type detection and file size validation. Supports both binary files (images, PDFs, archives) and text files, with automatic MIME type detection from file extension or content inspection. Attachment retrieval returns download URLs and metadata (filename, size, upload date, uploader), enabling AI agents to attach generated artifacts (reports, images, documents) to issues without manual file handling.

Exposes tools for querying user information, managing user assignments to issues, and checking permissions for specific operations. Implements role-based access control (RBAC) queries that determine if a user has permission to perform an action (edit issue, create page, etc.) without attempting the operation. User queries return user metadata (name, email, avatar, active status) and can filter by project or issue context, enabling AI agents to assign issues to appropriate team members and validate permissions before attempting operations.

Provides tools for querying Confluence spaces and Jira projects, including space/project metadata (name, key, description, avatar), configuration (permissions, issue types, custom fields), and member lists. Implements hierarchical space navigation (space → pages → children) and project-specific field discovery (custom fields, issue types, workflows), enabling AI agents to understand the structure of Confluence/Jira instances and adapt operations based on project-specific constraints.

Implements a dependency injection (DI) system using Python context managers and async context managers to provide JiraClient and ConfluenceClient instances to tool handlers, with per-request context isolation for multi-tenant deployments. Uses MainAppContext to store shared configuration (base URLs, authentication method) and per-request context to store user-specific credentials (from HTTP headers), enabling multiple users to authenticate with different credentials through the same server instance without credential leakage or cross-contamination.

Exposes 27+ Confluence tools for creating, reading, updating, and deleting pages within hierarchical space structures, with support for parent-child page relationships and content versioning. Uses the Confluence REST API v2 (Cloud) or v1 (Server/DC) with automatic content format adaptation between storage format (XHTML-like) and view format (rendered HTML), enabling AI agents to work with human-readable content while preserving Jira markup and embedded resources.

Implements the Model Context Protocol (MCP) server specification across three transport modes: stdio (for IDE integration), Server-Sent Events (SSE, for browser-based clients), and streamable-http (for service deployments). Uses FastMCP framework with async/await patterns and dependency injection to mount separate Jira and Confluence MCP instances on a shared AtlassianMCP parent server, enabling a single deployment to serve multiple AI platforms (Claude, Cursor, custom LLM clients) through their preferred transport mechanism.

Supports four authentication methods (API tokens, Personal Access Tokens, OAuth 2.0 3LO, bring-your-own-token) with environment variable precedence and per-request HTTP header overrides for multi-tenant deployments. Uses a configuration cascade that checks environment variables first, then HTTP headers (for service deployments), enabling both single-tenant (env-based) and multi-tenant (header-based) authentication patterns without code changes. OAuth 2.0 3LO flow is handled through a callback-based mechanism that exchanges authorization codes for access tokens, with automatic token refresh.

Exposes tools for managing Jira Agile boards, sprints, and backlogs, including sprint creation, issue assignment to sprints, board state queries, and workflow transitions. Implements board-specific queries that retrieve sprint metadata (start date, end date, goal), issue assignments, and board configuration (columns, swimlanes) through the Jira Agile API, with automatic handling of Scrum vs Kanban board differences and sprint state transitions (planning → active → closed).

Implements bidirectional content format conversion between Confluence storage format (XHTML-like), view format (rendered HTML), and markdown for AI consumption. Uses server-side rendering APIs (Confluence Content Transformation API) to convert between formats, preserving embedded resources, links, and markup while adapting content for different consumption patterns (AI-readable markdown vs browser-rendered HTML vs storage format for updates).

Implements MCP tool registration through FastMCP's decorator-based system, where each Jira/Confluence operation is registered as a callable tool with JSON schema describing inputs, outputs, and constraints. Uses dependency injection to provide JiraClient and ConfluenceClient instances to tool handlers, with automatic schema generation from Python type hints and docstrings. Tool discovery is handled through the MCP protocol's list_tools endpoint, enabling AI platforms to discover available operations and their signatures at runtime without hardcoded tool lists.

Provides tools for creating, updating, and querying issue worklogs (time spent on tasks) with automatic duration calculation from start/end times or explicit duration values. Supports worklog visibility settings (private, team, public) and automatic timestamp handling (created/updated dates), enabling AI agents to track time spent on issues without manual date/time formatting. Worklogs are stored per-issue and can be queried individually or aggregated across issues for reporting.

Exposes tools for creating, updating, and querying issue links (relationships between issues) with automatic link type validation and bidirectional relationship handling. Supports standard link types (blocks, relates to, duplicates, etc.) and custom link types defined in the Jira instance, with automatic enforcement of link type constraints (e.g., preventing circular dependencies in 'blocks' relationships). Link queries return both outbound (this issue blocks X) and inbound (Y blocks this issue) relationships, enabling AI agents to understand issue dependencies without manual relationship traversal.