**Short version:** Atlassian runs an official remote MCP server. Point your AI tool at `https://mcp.atlassian.com/v2/mcp`, sign in with OAuth, and the assistant can search, create and update Jira issues and read Confluence pages, strictly within your own permissions. It is Cloud-only; Data Center teams need a community server instead. Details below are from the atlassian-mcp-server README, checked in September 2026.

Setup by client
- Claude Code:
claude mcp add --transport http atlassian https://mcp.atlassian.com/v2/mcp, then/mcpto authenticate. - Codex:
codex mcp add atlassian --url https://mcp.atlassian.com/v2/mcp. - VS Code with Copilot: search
@mcp Atlassianin the Extensions view and install. - Cursor: add the Atlassian plugin from the Cursor marketplace.
What it covers
Jira, Confluence, Jira Service Management, Bitbucket, Compass and Loom on Atlassian Cloud. Typical tasks: search bugs, create and update issues, bulk-create work items from a spec, summarise Confluence documentation, and link content across products.
Admin settings that trip people up

- Cloud only.
- Data Center and Server sites are not supported.
- Delete and admin tools are off by default.
- An administrator must enable them.
- IP allowlisting still applies.
- Requests through the MCP server must come from an allowed IP.
- API tokens must be enabled for JSM.
- Jira Service Management tools only work with API token authentication.
On Jira Data Center?
Use the community sooperset/mcp-atlassian server, which supports on-premises deployments. Review what it can reach before connecting it to production.
A workflow that pays for itself
Ask the agent to read a bug report, reproduce it in the codebase, and file a Jira issue with the stack trace and a proposed fix, linked to the Confluence spec it came from. That is three tabs of copy-paste replaced by one prompt. See the Claude Code MCP guide for scoping this server to one project with --scope project.