Kiro + Atlassian MCP (Jira & Confluence)

The Atlassian MCP server lets Kiro read from Jira (issues, sprints, epics) and Confluence (pages, spaces) while you work. You can reference a Jira ticket by ID and Kiro will fetch its full description, acceptance criteria, and comments before writing the spec or code.

What You Can Do

“Implement the feature described in PROJ-1234” — Kiro reads the ticket first
“Write unit tests based on the acceptance criteria in PROJ-1234”
“Look up what the PROJ-89 epic covers and create sub-tasks for it”
“Summarise the Confluence page at /display/ARCH/API+Design+Guidelines”
“Generate a migration script matching the schema described in Confluence”

Prerequisites

Atlassian Cloud account (Jira/Confluence)
API token — generate at id.atlassian.com/manage-profile/security/api-tokens
Your Atlassian site URL (e.g. https://your-org.atlassian.net)

Installation

The Atlassian MCP server is published as an npm package:

# Install globally (optional — npx works without pre-installing)
npm install -g @atlassian/mcp-atlassian

Configuration

Add to .kiro/mcp.json in your project root:

{
  "mcpServers": {
    "atlassian": {
      "command": "npx",
      "args": ["-y", "@atlassian/mcp-atlassian"],
      "env": {
        "JIRA_URL": "https://your-org.atlassian.net",
        "JIRA_EMAIL": "${JIRA_EMAIL}",
        "JIRA_API_TOKEN": "${JIRA_API_TOKEN}",
        "CONFLUENCE_URL": "https://your-org.atlassian.net/wiki",
        "CONFLUENCE_EMAIL": "${JIRA_EMAIL}",
        "CONFLUENCE_API_TOKEN": "${JIRA_API_TOKEN}"
      }
    }
  }
}

Set the environment variables in your shell (not in the JSON file):

# Add to ~/.bashrc, ~/.zshrc, or Windows environment variables
export JIRA_EMAIL="you@yourcompany.com"
export JIRA_API_TOKEN="your_api_token_here"

On Windows (PowerShell):

$env:JIRA_EMAIL = "you@yourcompany.com"
$env:JIRA_API_TOKEN = "your_api_token_here"

What Kiro Can Access

Jira

Capability	Example prompt
Read issue by key	”What does PROJ-1234 require?”
List issues in sprint	”What’s in the current sprint for PROJ?”
List issues in epic	”Show me all stories under PROJ-100”
Search issues with JQL	”Find all open bugs assigned to me”
Read issue comments	”What was discussed in PROJ-1234?”

Confluence

Capability	Example prompt
Read page by title	”Read the API Design Guidelines page”
Search pages	”Find Confluence pages about authentication”
Read page content	”Summarise the data model described in Confluence”

Practical Examples

Feature implementation from a ticket:

User: Implement PROJ-456

Kiro:
1. Fetches PROJ-456 from Jira
2. Reads title: "Add email verification to registration flow"
3. Reads acceptance criteria, links to related tickets
4. Generates spec covering all acceptance criteria
5. Implements code with tests

Referencing architecture docs:

User: Create a new service following our architecture guidelines (see Confluence: "Service Template")

Kiro:
1. Fetches the Confluence page
2. Reads the agreed service structure
3. Generates new service following that template

Permissions Required

When creating the Atlassian API token, it uses your user’s permissions — Kiro will have the same read access you have in Jira and Confluence. The MCP server is read-only by default — it does not create or update Jira issues.

Troubleshooting

“Authentication failed”
Check that JIRA_EMAIL matches the email on your Atlassian account, and the API token is for that account.

“Project not found”
Verify the project key (e.g. PROJ) and that your account has access to it.

MCP server fails to start
Run npx @atlassian/mcp-atlassian manually in the terminal to see the error output directly.