Connect Claude to Atlassian: Jira and Confluence Options

Connecting Claude to Atlassian tools such as Jira and Confluence is possible through MCP servers that wrap Atlassian's REST API. There is no single official Anthropic connector for Atlassian, but several well-maintained options exist. fenbs can hold the shared decisions and next steps that matter to your team across sessions, so any assistant, not just the one you used today, can start from the same understanding.

The right connection method depends on whether you use Atlassian Cloud or Data Center, which API endpoints you need, and what your organization's access policy permits.

How the Atlassian REST API enables connection

Both Jira and Confluence expose REST APIs that allow authorized applications to read and write data. An MCP server wraps these APIs so a Claude client can call them within a session. You authenticate with an Atlassian API token, which you create in your Atlassian account settings and pass to the MCP server. Claude then queries issues, pages, or spaces based on what you ask.

Atlassian Cloud and Atlassian Data Center use different API endpoints and authentication flows. If your organization uses Data Center (self-hosted), check whether the MCP server you choose supports that deployment type explicitly.

Choosing an Atlassian MCP server

Before choosing a server, check a few things: does it use Atlassian's official REST API? Does it store any data it retrieves, or does it pass responses directly to the calling client? Is it actively maintained and open source so you can review what it does? Does it request only the API scopes your task needs?

An MCP server that requests read-only access to a specific project is more appropriate for most tasks than one that requests full Jira administration permissions. Set up the API token with the minimum permissions required and revoke it if you stop using the connection.

The Claude connectors explained guide describes the general MCP connection model. For comparison with other connector types, see Claude connectors and plugins.

What Claude can do with Jira access

With an authorized Jira MCP connection, Claude can help you search for issues, summarize a sprint's open tickets, draft descriptions for new issues, or analyze patterns across a project board. The quality of that help depends on the structure and completeness of your Jira data.

Claude cannot make decisions about which issues to prioritize, close, or assign without your input. Keep a person in the approval loop for any Jira write operation.

What Claude can do with Confluence access

With Confluence access, Claude can read page content, summarize documentation, help draft new pages, or answer questions about a space. This is useful for onboarding research, requirements work, or keeping a project brief current.

Confirm which spaces the API token grants access to. In large organizations, a token with full read access may return more content than you intend. Create a token scoped to the spaces you actually need.

Organizational access and approval

In many organizations, Atlassian is managed centrally. Before creating API tokens or installing external tools, check whether your IT or security team requires approval for third-party AI integrations. Some organizations block external API access to Jira and Confluence entirely until the integration is reviewed.

If you hit an access error when generating an API token, it may be an organization policy restriction. Contact your Atlassian administrator to understand what integrations are permitted.

Troubleshooting common Atlassian connection errors

Authentication errors are the most common issue when setting up an Atlassian MCP server. A 401 response means the API token is not being sent or is invalid. Check that the token is set in the environment variable the server expects, that it has not expired, and that the email address associated with the token matches the one you used to create it.

A 403 response means the token is valid but the account does not have permission to perform the requested operation on the specific project or space. Review the project or space permissions in Jira or Confluence for the account that owns the token. The token needs at least Browse Projects permission for Jira and Space: Read permission for Confluence.

If requests succeed for some projects but fail for others, the issue is almost always project-level permissions rather than the token itself. Add the service account to the relevant project role in Jira settings or the relevant space permissions in Confluence.

How to do this with fenbs

1. Create a free account at fenbs. Add a project for the work that uses Atlassian data.
2. Write the key decisions and rules that govern this project as Decisions and rules. These persist across sessions and are shown to any connected assistant first.
3. Connect Claude over MCP from Integrations and also set up your chosen Atlassian MCP server in the same Claude client.
4. At the start of a session, ask Claude to read the fenbs context first, then query Jira or Confluence for the current state. Combine both sources to answer your question.
5. Before ending the session, save Where we left off in fenbs with the next Jira tasks to check and any open decisions from the Confluence review.

fenbs does not connect to Atlassian. Its role is to keep the team's decisions, rules, and progress outside any one tool so handoffs between assistants and people stay coherent. For research across multiple connector types, see Claude research connectors.