How Atlassian and Anthropic connect Jira, Confluence, and Claude

Atlassian and Anthropic offer five ways to connect Claude to Jira and Confluence: Atlassian Skills for Claude, the Atlassian MCP server, AI deeplinks to Claude Code, Claude Agent for Jira, and Claude model selection in Rovo Agents.

To solve complex problems, Claude requires a complete view of how work, teams, and decisions connect across your organization. Atlassian’s Teamwork Graph provides that holistic map, allowing Claude to reason with the depth of a teammate who attended every planning session and read every spec—enriching the shared graph with every action it takes.

The end-to-end experience

Here’s how the pieces fit into a single workflow. Each step draws on one or more of the connection points detailed below.

  1. Plan: Capture the goal, requirements, and decisions in Jira and Confluence.

  2. Connect context: Add Atlassian Skills to Claude or connect through the Atlassian MCP Server.

  3. Delegate or build: Open the work in Claude Code, or assign a Jira work item to Claude Agent for Jira.

  4. Review: Review Claude’s proposed action or the draft pull request created from the Jira work.

  5. Keep work aligned: Keep activity and agent progress visible from the original Jira work item.

One connected loop: Jira and Confluence provide the intent and organizational context → Claude reasons and acts → people review the result → the work returns to Jira and Confluence as the shared record, ready to inform the next task.

Explore the ways to connect Claude and Atlassian

Atlassian MCP Server: Connect Claude through an open, permission-aware protocol

Best for: teams using Claude, Claude Desktop, Claude Code, or another MCP-compatible client that need access to Jira, Confluence, and Compass context and supported actions.

The Atlassian MCP Server connects compatible Claude clients to Atlassian Cloud through the open Model Context Protocol. Users authenticate through OAuth 2.1, and the server respects their existing Jira, Confluence, and Compass permissions. Organizations can apply administrator controls, review connected-app access, and monitor key activity.

Use Atlassian MCP to create Jira work items - all within Claude

Value moment: Claude can work with live Atlassian context without users manually copying tickets and pages into the prompt.

Requirements: Atlassian Cloud access, a supported MCP client, and the product permissions required for each requested read or write action. First-time connection may require site-administrator approval.

Get started with the Atlassian MCP server

Claude Agent for Jira: Assign Jira work to a Claude coding agent

Best for: software teams that want an AI coding agent to turn selected Jira work into a reviewable GitHub pull request.

Claude Agent for Jira is an Atlassian-built Forge app powered by Claude Managed Agents. From a Jira work item, users can assign the issue to Claude or mention the agent in a comment. The agent reads the Jira summary, description, and repository link, writes code, pushes a branch, and opens a draft pull request in GitHub while progress remains visible in Jira.

Status: Open beta. Availability and terms may change.

Let Claude Agent for Jira do the work for you so you can focus on what matters

Value moment: the work moves from a Jira assignment to human-reviewable code while the original issue remains the control point.

Requirements: Jira Cloud admin access, an Anthropic account with Claude Managed Agents and a customer-provided API key, plus a GitHub organization and service-account token with the required repository and pull-request permissions. The app is available for Jira Software and business projects, not Jira Service Management.

Data note: Claude Managed Agents is stateful and does not support Zero Data Retention. Anthropic’s terms govern its processing and retention, so organizations with ZDR requirements should not use this beta.

Explore Claude Agent for Jira on Atlassian Marketplace · Read the launch blog

AI coding tool deeplinks: Open Jira work directly in Claude Code

Best for: developers who want to move from a well-defined Jira work item into their coding agent without rebuilding the prompt.

Start from the Jira work item that already contains the goal, requirements, and acceptance criteria, then open it directly in Claude Code. The deeplink shortens the handoff from planning to implementation while keeping Jira as the source of intent.

Example:

  1. Open a ready-to-build Jira work item.

  2. Select the option to open the work in Claude Code

  3. Watch Claude Code receiving the Jira context and preparing an implementation plan.

  4. Review the plan before allowing code changes.

Value moment: the developer begins with the team’s agreed intent instead of reconstructing requirements in a separate AI prompt.

Requirements: Jira Cloud, Claude Code installed and authenticated, and access to the relevant repository and Jira work.

Learn about AI coding-tool deeplinks from Jira

Atlassian Skills: Add Jira and Confluence capabilities to Claude

Best for: teams that want a guided, low-setup way to use Atlassian knowledge and actions from Claude.

Install Atlassian Skills from Claude’s Skills directory to bring guided Jira and Confluence workflows into the Claude experience. Skills help Claude understand how to perform common Atlassian tasks, so users can move from a request to the right Jira or Confluence action without building a custom integration.

Open the Skills directory in Claude and select the Atlassian Jira and Confluence Skills.

  1. Authorize the Atlassian connection using the customer’s existing account.

  2. Ask: “Summarize the payments launch decision in Confluence and show the open Jira work.”

  3. Ask Claude to prepare a guided Jira or Confluence action, then review it before it runs.

Value moment: Claude moves from a general conversation to a workflow grounded in the team’s actual work and knowledge.

Requirements: an eligible Claude experience, access to the relevant Atlassian site, and any administrator approval required by organizational policy.

Learn about Claude Skills and the directory

Model selection: Choose Claude for a Rovo Agent

Best for: Rovo agent builders who want to match the model family to the reasoning and workflow needs of an agent.

Rovo Agents support model selection, including Anthropic Claude alongside other supported providers. Builders can choose the available model that best fits the task while keeping the agent grounded in Atlassian context, instructions, and permissions.

Choose OpenAI models to build Rovo Agents

  1. Open an eligible agent in Rovo Studio.

  2. Select an available Claude model for the agent.

  3. Test the agent on a reasoning task that uses Jira and Confluence context.

  4. Review the result and publish the agent for its intended audience.

Value moment: organizations get model choice without giving up the Atlassian context and governance that make the agent useful at work.

Requirements: an eligible Rovo plan and permission to create or configure Rovo Agents. Available model options can vary by plan, region, and product configuration.

See the model-selection announcement

How the experiences work together

Start with Atlassian Skills for the lowest-setup, guided experience; use the Atlassian MCP Server when you need an open, permission-aware protocol; use AI coding tool deeplinks to move a Jira work item straight into Claude Code; use Claude Agent for Jira to delegate an issue to a Claude coding agent; and use model selection in Rovo Agents to run Claude inside an Atlassian-built agent.⁠

If you want to…

Start with…

Add guided Jira and Confluence workflows to Claude

Atlassian Skills for Claude

Connect Claude or Claude Code through an open protocol

Atlassian MCP Server

Move a Jira issue into a developer’s coding flow

AI coding tool deeplinks to Claude Code

Delegate a Jira item to a Claude coding agent

Claude Agent for Jira

Use Claude as the model behind an Atlassian agent

Model selection in Rovo Agents

These experiences are complementary. MCP supplies the open connection layer, Claude Agent for Jira delegates implementation, AI coding tool deeplinks support developer handoffs, Skills provide a guided Claude experience, and Rovo model selection brings Claude into Atlassian-built agents.

Trust and control are part of the workflow

  • Enterprise-managed authentication: organizations can connect Claude to Atlassian with centralized identity, audit trails, and administrator governance, so employees work through their approved Claude experience while admins keep visibility and control over the connection. Learn about enterprise-managed authentication for Claude

  • Existing permissions apply: Atlassian MCP Server requests respect the user’s Jira, Confluence, and Compass permissions.

  • Secure authorization: supported clients use browser-based OAuth 2.1 rather than asking users to paste Atlassian passwords into Claude.

  • Administrator controls: organizations can manage connected-app access, domains, policies, and available audit information.

  • Human review: review high-impact Jira, Confluence, code, and pull-request actions before they are completed.

  • Client terms matter: when Atlassian content is sent to Claude, the organization’s Anthropic plan and terms govern Claude-side processing. Claude Agent for Jira is stateful and does not support Zero Data Retention, so teams with ZDR requirements should not use that beta.

FAQs

  • Does Claude respect my Jira and Confluence permissions? Yes. Every connection runs under your existing Jira, Confluence, and Compass permissions — Claude can only access what you already can, and high-impact actions can be reviewed before they run.

  • What’s the difference between Atlassian Skills and the Atlassian MCP server? Atlassian Skills is the guided, low-setup option: install it from Claude’s Skills directory to bring ready-made Jira and Confluence workflows into Claude. The Atlassian MCP Server is the open connection layer — it links any MCP-compatible client (Claude, Claude Desktop, Claude Code) to Atlassian Cloud through the Model Context Protocol, with OAuth 2.1 and admin controls.

  • Does Claude Agent for Jira work with Jira Service Management? No. Claude Agent for Jira supports Jira Software and Jira business projects, not Jira Service Management.

  • Does the Atlassian MCP server work with Claude Code? Yes. Claude Code is one of the MCP-compatible clients the Atlassian MCP Server connects to Atlassian Cloud.

  • Is Atlassian data retained by Anthropic? It depends on your Anthropic plan and terms, which govern any Claude-side processing of Atlassian content. One exception: Claude Agent for Jira runs on Claude Managed Agents, which is stateful and does not support Zero Data Retention — teams with ZDR requirements shouldn’t use that beta.

Getting started

  1. Choose the surface. Start in Claude with Skills, connect through the Atlassian MCP Server, use a Jira deeplink, or install the Claude Agent for Jira.

  2. Confirm access and terms. Review Atlassian permissions, Anthropic entitlements, data handling, and administrator policies.

  3. Begin with a bounded workflow. Use a defined Jira item, linked Confluence page, and test repository so people can verify the result.

  4. Keep humans in control. Review generated actions and code through existing Jira, Confluence, and GitHub governance.

Resources