Building a Jira AI Agent: Ticket Creation, Updates, and Sprint Management

Why Build AI Agents for Jira

Jira is the backbone of project tracking for software teams. An AI agent connected to Jira can automate ticket creation from Slack messages or emails, enrich tickets with context from codebases, estimate story points based on historical data, manage sprint planning, and generate sprint retrospective summaries — turning Jira from a manual data entry system into an intelligent project assistant.

Setting Up the Jira Client

Use API tokens for Jira Cloud authentication. The REST API provides comprehensive access to issues, boards, sprints, and workflows.

flowchart LR
    USER(["Customer"])
    CHANNEL{"Channel"}
    CHAT["Chat agent"]
    VOICE["Voice agent"]
    EMAIL["Email agent"]
    TRIAGE["Triage and<br/>intent detection"]
    KB[("Knowledge base<br/>RAG")]
    CRM[("CRM context")]
    AUTORES{"Auto resolvable?"}
    RESOLVE(["Resolved with<br/>cited answer"])
    HUMAN(["Tier 2 agent"])
    USER --> CHANNEL --> CHAT --> TRIAGE
    CHANNEL --> VOICE --> TRIAGE
    CHANNEL --> EMAIL --> TRIAGE
    TRIAGE --> KB
    TRIAGE --> CRM
    TRIAGE --> AUTORES
    AUTORES -->|Yes| RESOLVE
    AUTORES -->|No| HUMAN
    style TRIAGE fill:#4f46e5,stroke:#4338ca,color:#fff
    style AUTORES fill:#f59e0b,stroke:#d97706,color:#1f2937
    style RESOLVE fill:#059669,stroke:#047857,color:#fff
    style HUMAN fill:#0ea5e9,stroke:#0369a1,color:#fff

import httpx
from base64 import b64encode

class JiraClient:
    def __init__(self, domain: str, email: str, api_token: str):
        credentials = b64encode(
            f"{email}:{api_token}".encode()
        ).decode()
        self.http = httpx.AsyncClient(
            base_url=f"https://{domain}.atlassian.net/rest/api/3",
            headers={
                "Authorization": f"Basic {credentials}",
                "Content-Type": "application/json",
            },
            timeout=30.0,
        )

    async def create_issue(self, project_key: str, summary: str,
                           description: str, issue_type: str = "Task",
                           priority: str = "Medium",
                           labels: list[str] = None) -> dict:
        payload = {
            "fields": {
                "project": {"key": project_key},
                "summary": summary,
                "description": {
                    "type": "doc",
                    "version": 1,
                    "content": [
                        {
                            "type": "paragraph",
                            "content": [
                                {"type": "text", "text": description}
                            ],
                        }
                    ],
                },
                "issuetype": {"name": issue_type},
                "priority": {"name": priority},
            }
        }
        if labels:
            payload["fields"]["labels"] = labels

        response = await self.http.post("/issue", json=payload)
        response.raise_for_status()
        return response.json()

    async def search_issues(self, jql: str, max_results: int = 50) -> list:
        response = await self.http.post(
            "/search",
            json={
                "jql": jql,
                "maxResults": max_results,
                "fields": [
                    "summary", "status", "assignee",
                    "priority", "created", "updated",
                ],
            },
        )
        response.raise_for_status()
        return response.json()["issues"]

AI-Powered Ticket Creation

Let the agent parse unstructured requests — from Slack messages, emails, or voice transcripts — and create well-formatted Jira tickets.

async def create_ticket_from_request(
    jira: JiraClient,
    agent,
    raw_request: str,
    project_key: str,
):
    # Agent structures the raw input into Jira fields
    structured = await agent.run(
        prompt=(
            f"Parse this request into a Jira ticket.\n"
            f"Determine: summary (one line), description (detailed), "
            f"issue_type (Bug/Task/Story), priority (Highest/High/Medium/Low/Lowest), "
            f"and relevant labels.\n\n"
            f"Request: {raw_request}"
        )
    )

    ticket = await jira.create_issue(
        project_key=project_key,
        summary=structured.summary,
        description=structured.description,
        issue_type=structured.issue_type,
        priority=structured.priority,
        labels=structured.labels,
    )

    return ticket["key"]

JQL Queries for Intelligent Context

JQL (Jira Query Language) gives your agent powerful search capabilities. Use it to gather context before making decisions.

async def get_sprint_health(jira: JiraClient, project_key: str) -> dict:
    # Find current sprint issues
    in_progress = await jira.search_issues(
        f'project = {project_key} AND sprint in openSprints() '
        f'AND status = "In Progress"'
    )
    done = await jira.search_issues(
        f'project = {project_key} AND sprint in openSprints() '
        f'AND status = "Done"'
    )
    todo = await jira.search_issues(
        f'project = {project_key} AND sprint in openSprints() '
        f'AND status = "To Do"'
    )
    blocked = await jira.search_issues(
        f'project = {project_key} AND sprint in openSprints() '
        f'AND status = "Blocked"'
    )

    return {
        "total": len(in_progress) + len(done) + len(todo) + len(blocked),
        "done": len(done),
        "in_progress": len(in_progress),
        "todo": len(todo),
        "blocked": len(blocked),
        "completion_pct": round(
            len(done) / max(len(in_progress) + len(done) + len(todo) + len(blocked), 1) * 100
        ),
    }

Workflow Transitions

Moving tickets through workflow states requires knowing the available transitions for the current status.

async def transition_issue(
    jira: JiraClient, issue_key: str, target_status: str
):
    # Get available transitions
    response = await jira.http.get(
        f"/issue/{issue_key}/transitions"
    )
    transitions = response.json()["transitions"]

    # Find the transition that leads to our target status
    transition = next(
        (t for t in transitions if t["to"]["name"] == target_status),
        None,
    )

    if not transition:
        available = [t["to"]["name"] for t in transitions]
        raise ValueError(
            f"Cannot transition to '{target_status}'. "
            f"Available: {available}"
        )

    await jira.http.post(
        f"/issue/{issue_key}/transitions",
        json={"transition": {"id": transition["id"]}},
    )

# Agent-driven bulk status update
async def close_stale_tickets(jira: JiraClient, project_key: str, agent):
    stale = await jira.search_issues(
        f'project = {project_key} AND status = "In Progress" '
        f'AND updated <= -14d'
    )

    for issue in stale:
        key = issue["key"]
        summary = issue["fields"]["summary"]

        decision = await agent.run(
            prompt=f"Ticket {key} ('{summary}') has not been updated in "
                   f"14 days. Should we move it to Blocked, close it, "
                   f"or leave it? Explain briefly."
        )

        if decision.action != "leave":
            await transition_issue(jira, key, decision.target_status)
            await jira.http.post(
                f"/issue/{key}/comment",
                json={"body": {
                    "type": "doc", "version": 1,
                    "content": [{"type": "paragraph", "content": [
                        {"type": "text", "text": f"AI Agent: {decision.reason}"}
                    ]}]
                }},
            )

FAQ

How do I handle Jira's Atlassian Document Format for descriptions?

Jira Cloud V3 API uses Atlassian Document Format (ADF), a JSON-based rich text format. Simple text wraps in paragraph nodes as shown above. For complex formatting (tables, code blocks, bullet lists), build nested ADF node structures. Consider writing a helper function that converts markdown to ADF to simplify agent output formatting.

What are the Jira API rate limits?

Jira Cloud allows roughly 100 requests per minute for basic plans and higher limits for premium. Implement rate limiting on your client side with a token bucket or semaphore. The API returns Retry-After headers on 429 responses — respect those values before retrying.

Can the AI agent assign tickets to specific team members?

Yes. Use the assignee field in the create or update payload with the user's Atlassian account ID. To find account IDs, query /rest/api/3/user/search?query=username. Your agent can learn team members' areas of expertise and intelligently assign based on ticket content and past assignments.

---

#Jira #ProjectManagement #RESTAPI #AIAgents #SprintManagement #AgenticAI #LearnAI #AIEngineering