Jira MCP Server Cursor IDE Setup (2025-2026)

How do you set up the Jira MCP server in Cursor? Generate an Atlassian API token, add a jira block to ~/.cursor/mcp.json with that token, your site URL, and your account email, then restart Cursor. That's it — no separate OAuth app to register for a personal setup. Once it's running, your AI can read tickets, pull acceptance criteria into a coding session, and transition issues, all without switching to a browser tab.

This config hasn't changed between the 2025 and 2026 Cursor releases — if you're following an older screenshot or a 2025-dated tutorial, the steps below still apply. This guide is scoped to Jira issue tracking: sprints, tickets, transitions, JQL. Two related guides cover adjacent ground — Confluence MCP server setup for docs and spaces, and the combined Atlassian bundle if you want Jira, Confluence, and Bitbucket behind one OAuth connection instead of a Jira-only token.

Quick Reference

| | |
|---|---|
| Auth method | Atlassian API token |
| Config keys | ATLASSIAN_API_TOKEN, ATLASSIAN_SITE_URL, ATLASSIAN_USER_EMAIL |
| Package | @atlassian/mcp-atlassian |
| Setup time | ~10 minutes |
| Covers | Jira tickets, sprints, transitions, JQL — not Confluence docs (see Confluence guide) |
| Team setup note | Use a dedicated service account token, not a shared personal one |

What You Can Do with Jira + Cursor MCP

Once connected:

The last one is especially useful during development — pull the full acceptance criteria into your Cursor context before writing code.

Prerequisites

Step 1: Get an Atlassian API Token

1. Go to id.atlassian.com/manage-profile/security/api-tokens
2. Click Create API token
3. Give it a name like "Cursor MCP"
4. Copy the token — you won't be able to see it again

Your Atlassian site URL looks like: https://yourcompany.atlassian.net

Step 2: Add to Cursor MCP Config

Open ~/.cursor/mcp.json (or press Cmd/Ctrl + Shift + P → "Open MCP Settings"):

{
  "mcpServers": {
    "jira": {
      "command": "npx",
      "args": ["-y", "@atlassian/mcp-atlassian"],
      "env": {
        "ATLASSIAN_API_TOKEN": "your_api_token_here",
        "ATLASSIAN_SITE_URL": "https://yourcompany.atlassian.net",
        "ATLASSIAN_USER_EMAIL": "you@yourcompany.com"
      }
    }
  }
}

Replace the values with your actual token, site URL, and email.

Note: If multiple teammates need this connection, don't share one person's API token. Atlassian tokens are tied to an individual account and inherit that account's exact permissions — including whatever it can't see. For a team setup, create a dedicated service account in Jira, scope its project access deliberately, and generate the token from that account instead.

Step 3: Restart Cursor

Quit Cursor completely and reopen it. Check that it loaded: View → Output → MCP

Step 4: Test Your Connection

In Cursor chat, try:

What Jira issues are assigned to me?

Or pull context for an issue you're working on:

Show me the full description and acceptance criteria for PROJ-123

Practical Workflows

Start Your Day

Show me all Jira issues assigned to me that are In Progress or In Review

This gives you a quick standup summary without opening Jira.

Code with Ticket Context

Before implementing a feature, paste the ticket number in Cursor:

Pull the requirements from PROJ-456 and then help me design the database schema for it

Cursor will fetch the full ticket description and use it to inform the implementation.

Find Stale Issues

List all tickets in the current sprint that haven't been updated in more than 2 days

Pre-PR Checklist

Get the acceptance criteria from PROJ-789 and check my current code changes against them

Sprint Retro Prep

List every ticket completed in the last sprint, group them by whether they were bugs or features, and flag any that got reopened after being marked Done

Useful for retro discussions — reopened tickets are usually a better signal of process problems than raw velocity numbers.

Troubleshooting

"Authentication failed" or 401 errors

"Project not found" errors

Slow responses

Server not appearing in Cursor

Config from an older tutorial doesn't seem to load
If you copied a jira block from a 2025-dated guide or a forum post, the field names above (ATLASSIAN_API_TOKEN, ATLASSIAN_SITE_URL, ATLASSIAN_USER_EMAIL) are still current — the most common break is a leftover key from a different package version sitting alongside them. Delete the whole block and retype it fresh rather than patching an old one.

Considering Atlassian's official OAuth remote server instead of this token setup?
Worth knowing before you switch: Atlassian's hosted Rovo MCP Server (mcp.atlassian.com) went through an endpoint migration in 2026 — the older /v1/sse endpoint stopped being served after June 30, 2026, replaced by /v1/mcp/authv2, alongside a move to a new OAuth identity provider for dynamic client registration starting May 27, 2026. If you're following a guide or a saved config from earlier in 2026 that still points at /v1/sse, that's why it stopped connecting — it's not a token or permissions problem. The API-token method in this guide is unaffected by that migration since it doesn't touch the OAuth endpoint at all. See the Atlassian MCP bundle guide for the current OAuth setup if you'd rather use one connection across Jira, Confluence, and Bitbucket instead of this Jira-only token.

Jira-Specific Tips

Use project keys — Instead of "the backend project," say "the BACKEND project" (using the Jira project key). This avoids ambiguity.

JQL works — You can ask Cursor to filter by JQL: "Find Jira issues matching: project = BACKEND AND status = 'In Progress' AND assignee = currentUser()"

Combining with code — The real power is combining ticket context with your code. Try: "I'm implementing the feature described in PROJ-456 — review my current implementation and tell me if I'm missing anything from the acceptance criteria"

Looking for Confluence, Not Jira?

Landed here searching for "confluence mcp server cursor ide setup 2026"? You're one page off — this guide covers Jira ticket tracking (issues, sprints, transitions, JQL). Confluence runs on a different REST API with different tools, and the dedicated Confluence MCP Server Cursor IDE Setup guide covers it end to end: the confluence-mcp-server package, the CONFLUENCE_BASE_URL / CONFLUENCE_USERNAME / CONFLUENCE_API_TOKEN config block, space scoping, and the ADF content-format gotcha that trips people up on their first write. The same Atlassian API token from Step 1 above works for both products — you're generating one credential, not two — but Jira and Confluence need separate mcpServers entries in mcp.json since the packages and env variable names aren't interchangeable.

If your team runs both, don't try to extend the jira block above with Confluence fields; add a second, independent confluence entry alongside it, following the exact JSON block on the dedicated guide.

Frequently Asked Questions

Q: Does this setup also give me Confluence access, or do I need a separate server?
A: Technically yes — @atlassian/mcp-atlassian exposes both under one token. But this guide only documents the Jira tools. If docs and spaces are your actual goal, use the dedicated Confluence guide instead; it covers the doc-specific workflows this page skips.

Q: Can I pull a Jira ticket's linked Confluence spec into a coding session?
A: Yes — ask something like "Get the acceptance criteria from PROJ-456, then find the linked design doc and summarize it." The same Atlassian connection handles both lookups without extra config.

Q: Does this work with Jira Server or Data Center, not just Jira Cloud?
A: Officially, support is Cloud-first. Server/Data Center instances behind a VPN or custom auth setup often work but aren't guaranteed — if your org runs an on-prem Jira instance, test the connection with a read-only query before relying on it for anything write-related.

Q: My AI can see issues but can't transition or comment on them — why?
A: Check the permission scope on the Atlassian account tied to your API token, not the token itself. The MCP server inherits whatever that account can do in the Jira UI — if it can't move a ticket from "In Review" to "Done" manually, the AI can't either.

Q: If my team's repos live in GitLab or Bitbucket instead of GitHub, does this still work the same way?
A: Yes — the Jira MCP connection is independent of where your code lives. Pair it with the GitLab MCP server or Bitbucket MCP server to get the same "pull ticket, check code, create branch" workflow regardless of which platform hosts your repos.

Q: Is the Jira setup different for Cursor's 2025 vs. 2026 releases?
A: No — the mcp.json schema and the @atlassian/mcp-atlassian env variables in this guide haven't changed across recent Cursor versions.

Q: Can Cursor bulk-transition multiple Jira tickets in one request?
A: It can work through a list, but scope the list explicitly — ask "Move PROJ-101, PROJ-104, and PROJ-110 to Done" rather than "move everything in the sprint to Done." There's no true bulk-transition endpoint in the Jira REST API, so the MCP server loops through issues one at a time; an unscoped request is more likely to hit a rate limit or transition something you didn't mean to touch.

Q: Does this respect Jira's custom fields and screen schemes, or only default fields?
A: It reads and writes whatever fields the Jira REST API exposes for your project, custom fields included. But if a field is hidden by a screen scheme for the transition you're attempting, the API rejects the write the same way the Jira UI would. If an update to a custom field silently fails, check whether that field is actually on the screen for the target status first.

Q: Can the Jira MCP server create new issues, or only read and transition existing ones?
A: Creation is supported — ask something like "Create a bug in the BACKEND project titled 'Checkout button unresponsive on Safari' with a description of the repro steps" and it works the same way transitioning a ticket does, subject to the same account permissions. The practical limit isn't the tool, it's the target project's required fields: if BACKEND has a mandatory custom field with no default, a creation request that omits it fails the same way it would from the Jira UI, so ask Cursor to check the project's create-screen fields first if a new-issue prompt keeps getting rejected.

Q: Does this see Advanced Roadmaps data — epics and cross-project dependencies — or just single-project issues?
A: Basic epic-to-story links inside one project read and write fine, since those are ordinary issue fields. Advanced Roadmaps-specific concepts — cross-project epic rollups, capacity planning views, dependency graphs across multiple boards — sit in a separate Jira Premium feature with its own API surface that @atlassian/mcp-atlassian doesn't cover as of this guide. If your team relies on Advanced Roadmaps for planning, treat this MCP connection as ticket-level (single issues, single-project epics), not a substitute for the roadmap view itself.

Q: What's the exact atlassian jira mcp server cursor ide setup for 2025 2026, step by step?
A: Generate an Atlassian API token at id.atlassian.com (Step 1), add the jira block to ~/.cursor/mcp.json with ATLASSIAN_API_TOKEN, ATLASSIAN_SITE_URL, and ATLASSIAN_USER_EMAIL (Step 2), restart Cursor (Step 3), then confirm with a prompt like "what Jira issues are assigned to me?" (Step 4). The config hasn't changed between Cursor's 2025 and 2026 release lines, so a 2025-dated tutorial's field names still apply.

Q: I only searched "jira mcp server cursor ide setup 2026" — is this the complete guide or just an overview?
A: This is the complete guide — token generation, the full mcp.json block, verification steps, practical workflows, and troubleshooting for actual errors people hit (auth failures, project-not-found, stale configs from older tutorials). If you specifically need Confluence docs instead of Jira tickets, use the dedicated Confluence guide — the token generation step is identical, but the tools and config are different.

Q: Did anything about Jira MCP setup actually change for 2026, or is "2025-2026" just a formality in the title?
A: One real change, but it's on the OAuth side, not this guide's token method: Atlassian's official remote server retired its /v1/sse endpoint after June 30, 2026, in favor of /v1/mcp/authv2, as part of a broader move to a new OAuth identity provider starting May 27, 2026. The API-token setup in this guide never touched that endpoint, so it's unaffected either way — but if you're troubleshooting a stalled OAuth-based connection from an early-2026 config, that migration is the likely cause.

---

Related Guides

---