SuperAPI AI Directory
Partner with us中文AI newsFavorites
All toolsAI codingmcp-server-atlassian-jira

mcp-server-atlassian-jira

Atlassian Jira Cloud: work with projects, issues, comments and related development information.

Open website

About this tool

Purpose

mcp-server-atlassian-jira is an MCP (Model Context Protocol) server implemented in Node.js/TypeScript, used to connect AI systems (LLMs) with Atlassian Jira Cloud. It enables MCP-supporting AI assistants such as Claude and Cursor AI to directly operate Jira's project management, issue tracking, and workflows, completing query and management tasks through natural language.

Main Capabilities

This server provides 5 general-purpose tools that can access any Jira REST API v3 endpoint, covering:

  • List/get projects (such as /rest/api/3/project/search, /rest/api/3/project/{projectKeyOrId})
  • Search/get issues using JQL or ID (/rest/api/3/search/jql, /rest/api/3/issue/{issueIdOrKey})
  • Create issues (POST /rest/api/3/issue)
  • Get/execute status transitions (/rest/api/3/issue/{issueIdOrKey}/transitions)
  • List/add/update/delete comments (/rest/api/3/issue/{issueIdOrKey}/comment)
  • List/add/update/delete worklogs (/rest/api/3/issue/{issueIdOrKey}/worklog)
  • View development information (commits, PRs)
  • User and status queries (/rest/api/3/myself, /rest/api/3/user/search, /rest/api/3/status, /rest/api/3/issuetype, /rest/api/3/priority)

By default, output uses the TOON (Token-Oriented Object Notation) format, which can reduce token consumption by 30-60% compared with JSON; it can also be switched to JSON via --output-format json or outputFormat: "json". All tools support optionalJMESPath (jq) filtering to extract specific data. For large responses exceeding 40k characters (about 10k tokens), they are automatically truncated with a notice, and the complete raw response is saved to /tmp/mcp/mcp-server-atlassian-jira/<timestamp>-<random>.txt.

Usage

  1. Obtain Jira credentials: go to Atlassian API Tokens to create an API token, and copy it immediately.
  2. Connect an AI assistant:
    • Claude Desktop users: add the configuration in the config file ~/.claude/claude_desktop_config.json, restart Claude Desktop, and the jira server will appear in the status bar.
    • Other AI assistants: install the server globally (see the official repository for the npm installation command), then configure the AI assistant to use the MCP server with STDIO transport.
    • Alternative: create ~/.mcp/configs.json for system-level configuration; the configuration key can use "jira", "atlassian-jira", "@aashari/mcp-server-atlassian-jira", or "mcp-server-atlassian-jira".
  3. Direct CLI access: CLI commands correspond to MCP tools and support --output-format (toon by default or json), --jq (JMESPath filtering), --query-params (query parameters as a JSON string).

Applicable Scenarios

  • Developers quickly obtaining issue details and development context
  • Project managers tracking progress, priority, and team assignments
  • Scrum Masters managing sprints and workflow statuses
  • Team leads monitoring project health and issue resolution
  • QA engineers tracking defects and test status
  • Any user who wants to interact with Jira through natural language

Limitations

  • Only supports Jira Cloud, not Jira Server/Data Center (may be added in future versions).
  • Requires Node.js 18.0.0 or later.
  • Requires an Atlassian account with the corresponding projectpermissions, and the API token must be valid.
  • Site name format: Jira URL is https://mycompany.atlassian.net the site name is mycompany.
  • Data security: The tool runs entirely locally, uses the user's own Jira credentials, does not send data to third parties, and only accesses the authorized scope.

Technical details

  • Version 3.2.1 (December 2025): Added TOON output format; implemented automatic response truncation; raw responses are saved to /tmp/mcp/mcp-server-atlassian-jira/; updated to MCP SDK v1.23.0; fixed the deprecated /rest/api/3/search endpoint (now uses /rest/api/3/search/jql); updated dependencies (Zod v4.1.13, Commander v14.0.2).
  • The architecture follows a 5-layer MCP architecture: CLI layer, Tools layer, Controllers layer, Services layer, Utils layer.
  • Debugging: Set the DEBUG environment variable to enable debug logs; logs are written to ~/.mcp/data/mcp-server-atlassian-jira.<session-id>.log.
  • Migrating from v2.x: v3.0 replaces 8+ specific tools with 5 generic HTTP method tools, for example jira_ls_projects becomes jira_get with path /rest/api/3/project/search.

FAQ

  • Authentication failure or 403: Check the API token permissions and expiration, verify the site name format, and use npx -y @aashari/mcp-server-atlassian-jira get --path "/rest/api/3/myself" to test credentials.
  • Resource not found or 404: Check the case of the API path, use the project key (such as DEV) instead of the project name, issue keys include the project prefix (such as DEV-123), and confirm access permissions.
  • Search returns no results: Try different search terms, use the project key, and check the JQL syntax.
  • Claude Desktop integration issues: Restart after updating the configuration, and confirm the configuration file location (macOS: ~/.claude/claude_desktop_config.json; Windows: %APPDATA%\Claude\claude_desktop_config.json).

For more help, visit the GitHub repository github.com/aashari/mcp-server-atlassian-jira or submit an Issue.

Related tools

Browse category