Atlassian MCP teaches Claude to use Jira and Confluence through an MCP server. It covers choosing between the hosted Atlassian server, the community mcp-atlassian server and a self-hosted setup, then configuring OAuth or tokens.
Queries are written in JQL for Jira and CQL for Confluence and are tested with a one-result probe first. Write and bulk operations wait for a read-only permission check and a confirmation.
What You Get
- MCP server configuration for Atlassian Cloud or a Server and Data Center setup
- JQL and CQL examples with explanations
- Tool call code with pagination and error recovery
- Authentication setup steps for OAuth and personal access tokens
- A note on the permission scopes the workflow needs
Common Use Cases
- Querying open Jira issues in a sprint with JQL
- Searching Confluence pages by space and keyword
- Syncing meeting notes into Jira tickets
- Automating sprint and backlog workflows
- Debugging an Atlassian MCP integration
What Makes It Unique
The skill is specific to Atlassian over MCP and gives concrete setup for both the hosted Rovo server and the community server. Safety rules are part of the method: probe with maxResults=1, run read-only first, back off on rate limits and confirm before changing production data.
Under The Hood
The skill is a SKILL.md with five reference files for server setup, Jira queries, Confluence operations, authentication and common workflows. It has no scripts of its own. The setup it describes registers an MCP server with the claude mcp add command, or runs the mcp-atlassian server through uvx, and that server then calls Jira and Confluence APIs with your credentials, so it reaches the network and can create or edit tickets and pages. The skill tells the agent to keep tokens in environment variables and to confirm before writes.