STARTER_CHARACTER = 🎟️

JIRA Conventions

Jira is an issue tracking sytem. Use the atl mcp, jira cli, and jira api with curl according to the directions for the entity and operation.

Instance

Use environment variables for JIRA configuration:

Variable	Description
`JIRA_URL`	Base URL (e.g., `https://mcghealth.atlassian.net/`)
`JIRA_EMAIL`	User email for authentication
`JIRA_API_TOKEN`	API token for authentication

•Browse URL: ${JIRA_URL}browse/{KEY}

MCP Tool Data Types

When calling tools, use correct data types for parameters:

Type	Example	Wrong
Integer	`limit=10`	`limit="10"`
String	`issue_key="DEVX-123"`	`issue_key=DEVX-123`
Object	`fields={"summary": "Title"}`	`fields="{\"summary\": \"Title\"}"`
Array	`labels=["bug", "urgent"]`	`labels="[\"bug\", \"urgent\"]"`

Incorrect types cause validation errors like: '10' is not of type 'integer'

Known bug: mcp__atl__jira_update_issue cannot set numeric custom fields (e.g., story points). The fields dict gets serialized to a string by pydantic, causing either a validation error or a Jira rejection. Use the curl approach for numeric custom fields (see Story Points section below).

Issue Key Recognition

{KEY} refers to the normalized JIRA issue key in format PROJ-### (e.g., DEVX-209).

The default jira project is DEVX. DEVX-###.

Accept flexible input formats:

•PROJ-209 (full key)
•proj-209 (lowercase)
•PROJ209 (no hyphen)
•proj209 (lowercase, no hyphen)
•209 (number only, requires project context)

Always normalize to uppercase with hyphen: DEVX-209

Writing Standards

•No emojis
•No glyphs
•No emdashes
•Direct, technical language
•Clear imperative sentences

Formatting Conventions

•Prose is normal paragraphs
•Code and commands use code blocks
•Citations from documentation or other people use block quotes

MCP ATL Markup Format

IMPORTANT: The MCP ATL tools (mcp__atl__jira_create_issue, mcp__atl__jira_update_issue, mcp__atl__jira_add_comment) convert Markdown to Jira wiki format automatically.

Use Markdown for content

Code blocks with shebangs:

code

#!/bin/bash
set -euo pipefail

Nested lists (use indentation):

•
Level 1
- •
  Level 2
  - •Level 3

Wiki markup panels are supported

Panels can wrap markdown content:

code

{panel:bgColor=#deebff}
### Context
Content here uses markdown.
{panel}

Do NOT use wiki markup for content

Passing wiki markup for regular content causes double-conversion bugs:

•{code}#!/bin/bash{code} → # becomes h1.
•** nested item → ** becomes __

Known Limitations

•Code blocks with # - Fixed in PR #894, pending merge. Using local fork at ~/tmp/mcp-atlassian until released.
•Jira does not support syntax highlighting in code blocks

Description Format

Use panel template for new issues (panel is wiki markup, content is markdown):

code

{panel:bgColor=#deebff}
### Context
<narrative: background, expected outcomes, why this matters>
{panel}

### The Work
<bullet points of tasks to complete>

### Details
<optional: technical information - config values, resources, commands, data>

Operations

Epic

create

code

mcp__atl__jira_create_issue(
  project_key="PROJ",
  issue_type="Epic",
  summary="Epic title",
  description="{panel:bgColor=#deebff}\n### Context\nBackground and why this matters.\n{panel}\n\n### The Work\n- First task\n- Second task\n\n### Details\nTechnical information."
)

read

code

mcp__atl__jira_get_issue(
  issue_key="PROJ-123"
)

update

code

mcp__atl__jira_update_issue(
  issue_key="PROJ-123",
  fields={
    "summary": "New title",
    "description": "{panel:bgColor=#deebff}\n### Context\nUpdated background.\n{panel}\n\n### The Work\n- Updated tasks\n\n### Details\nUpdated details."
  }
)

cancel

code

mcp__atl__jira_transition_issue(
  issue_key="PROJ-123",
  transition_id="91"
)

list children

code

jira issue list -q "parent = PROJ-123" -p PROJ --plain

Task

create

code

mcp__atl__jira_create_issue(
  project_key="PROJ",
  issue_type="Task",
  summary="Task title",
  description="{panel:bgColor=#deebff}\n### Context\nBackground and why this matters.\n{panel}\n\n### The Work\n- First task\n- Second task\n\n### Details\nTechnical information."
)

create with parent

code

mcp__atl__jira_create_issue(
  project="PROJ",
  issue_type="Task",
  summary="Task title",
  description="{panel:bgColor=#deebff}\n### Context\nBackground and why this matters.\n{panel}\n\n### The Work\n- First task\n- Second task\n\n### Details\nTechnical information.",
  additional_fields={"parent": "PROJ-123"}
)

read

code

mcp__atl__jira_get_issue(
  issue_key="PROJ-456"
)

update

code

mcp__atl__jira_update_issue(
  issue_key="PROJ-456",
  fields={
    "summary": "New title",
    "description": "{panel:bgColor=#deebff}\n### Context\nUpdated background.\n{panel}\n\n### The Work\n- Updated tasks\n\n### Details\nUpdated details."
  }
)

cancel

code

mcp__atl__jira_transition_issue(
  issue_key="PROJ-456",
  transition_id="91"
)

search

code

jira issue list -q "project = PROJ AND status != Done ORDER BY created DESC" -p PROJ --plain

Link

link to epic

code

mcp__atl__jira_link_to_epic(
  issue_key="PROJ-456",
  epic_key="PROJ-123"
)

create link

code

mcp__atl__jira_create_issue_link(
  inward_issue_key="PROJ-456",
  outward_issue_key="PROJ-789",
  link_type="Blocks"
)

Common link types: Blocks, Dependency, Duplicate, Relates, Parent/Child

get types

code

mcp__atl__jira_get_link_types()

remove

code

mcp__atl__jira_remove_issue_link(
  issue_key="PROJ-456",
  link_id="12345"
)

Assignment

assign

code

mcp__atl__jira_update_issue(
  issue_key="PROJ-456",
  fields={"assignee": "user@example.com"}
)

unassign

code

mcp__atl__jira_update_issue(
  issue_key="PROJ-456",
  fields={"assignee": null}
)

get user

code

mcp__atl__jira_get_user_profile()

Transition/Move

get available

code

mcp__atl__jira_get_transitions(
  issue_key="PROJ-456"
)

change status

code

mcp__atl__jira_transition_issue(
  issue_key="PROJ-456",
  transition_id="31"
)

Rule: When transitioning an issue to Done, always assign it to the current user (from JIRA_EMAIL) if unassigned. Use the curl approach for assignment since the MCP tool has serialization issues with the assignee field.

Comment

add

code

mcp__atl__jira_add_comment(
  issue_key="PROJ-456",
  comment="Comment text here."
)

read

code

mcp__atl__jira_get_issue(
  issue_key="PROJ-456",
  fields="summary,comment",
  comment_limit=10
)

Note: The comment field must be included in fields to retrieve comments. The comment_limit parameter only limits how many are returned.

update

code

curl -X PUT \
  -u "${JIRA_EMAIL}:${JIRA_API_TOKEN}" \
  -H "Content-Type: application/json" \
  -d '{"body":"Updated comment text."}' \
  "${JIRA_URL}rest/api/2/issue/PROJ-456/comment/12345"

delete

code

curl -X DELETE \
  -u "${JIRA_EMAIL}:${JIRA_API_TOKEN}" \
  "${JIRA_URL}rest/api/2/issue/PROJ-456/comment/12345"

Attachment

add

code

mcp__atl__jira_update_issue(
  issue_key="PROJ-456",
  fields={},
  attachments="/path/to/file.pdf"
)

read

code

mcp__atl__jira_get_issue(
  issue_key="PROJ-456",
  fields="attachment"
)

The attachment ID is in the response at attachments array.

download

code

mcp__atl__jira_download_attachments(
  issue_key="PROJ-456",
  download_path="/tmp/downloads"
)

delete

code

curl -X DELETE \
  -u "${JIRA_EMAIL}:${JIRA_API_TOKEN}" \
  "${JIRA_URL}rest/api/3/attachment/12345"

Priority

get

code

mcp__atl__jira_get_issue(
  issue_key="PROJ-456"
)

The priority is in the response at priority.name.

set

code

mcp__atl__jira_update_issue(
  issue_key="PROJ-456",
  fields={"priority": {"name": "High"}}
)

Valid values: Highest, High, Medium, Low, Lowest

Label

add

First get current labels, then update with the merged list.

code

mcp__atl__jira_get_issue(
  issue_key="PROJ-456"
)

code

mcp__atl__jira_update_issue(
  issue_key="PROJ-456",
  fields={"labels": ["existing-label", "new-label"]}
)

remove

First get current labels, then update with the filtered list.

code

mcp__atl__jira_get_issue(
  issue_key="PROJ-456"
)

code

mcp__atl__jira_update_issue(
  issue_key="PROJ-456",
  fields={"labels": ["remaining-label"]}
)

list

code

mcp__atl__jira_get_issue(
  issue_key="PROJ-456"
)

Labels are in the response at labels array.

Story Points

Custom field: customfield_10814 (Story point estimate)

The MCP ATL tool cannot set this field due to a numeric type serialization bug. Use curl instead.

set

code

curl -s -X PUT \
  -u "${JIRA_EMAIL}:${JIRA_API_TOKEN}" \
  -H "Content-Type: application/json" \
  -d '{"fields":{"customfield_10814":0.5}}' \
  "${JIRA_URL}rest/api/2/issue/PROJ-456"

set with assignee (bulk pattern)

code

for KEY in DEVX-101 DEVX-102 DEVX-103; do
  curl -s -X PUT \
    -u "${JIRA_EMAIL}:${JIRA_API_TOKEN}" \
    -H "Content-Type: application/json" \
    -d '{"fields":{"assignee":{"emailAddress":"user@example.com"},"customfield_10814":0.5}}' \
    "${JIRA_URL}rest/api/2/issue/${KEY}"
done

get

code

mcp__atl__jira_get_issue(
  issue_key="PROJ-456",
  fields="summary,customfield_10814"
)

The value is at customfield_10814 in the response. Common values: 0.5, 1, 2, 3, 5, 8, 13.