Atlassian introduces ‘always-on’ capabilities for agentic development workflows

The core of this announcement centers on the recognition that for AI to be truly effective in a professional software engineering environment, it must possess deep, organization-specific context. The new features—ranging from Code Context and Agent Context Controls to the DX for Agentic Development suite—are engineered to ensure that AI agents do not operate in a vacuum. Instead, they function as collaborative partners that understand the nuances of a company’s specific codebase, architectural standards, and project management history.

The Evolution of Agentic Workflows

The software development landscape has seen a rapid transition from basic AI-powered autocomplete features to autonomous "agents" capable of executing multi-step tasks. However, many engineering teams have struggled to integrate these agents into their daily operations due to concerns regarding security, lack of context, and the difficulty of measuring return on investment. Atlassian’s latest update aims to address these hurdles by embedding AI directly into the "Teamwork Graph"—the underlying data structure that connects the various tools used by development teams, including Jira, Confluence, and Bitbucket.

Taroon Mandhana, Atlassian’s CTO for AI & Teamwork, emphasized that the primary bottleneck in modern AI software engineering is not the sophistication of the Large Language Models (LLMs) themselves, but rather the lack of organizational context. "Enterprises need more than isolated sessions and one-off prompts," Mandhana noted. By utilizing Jira as a robust system of record, Atlassian is effectively creating a foundation where agents can be orchestrated safely, ensuring that every AI-generated contribution aligns with existing project backlogs and architectural guidelines.

Chronology of AI Integration at Atlassian

Atlassian’s journey toward agentic development has been a strategic, multi-year process. The company began its AI pivot by integrating Rovo, its intelligent search and discovery tool, into the Atlassian platform. This allowed users to query information across vast amounts of documentation in Confluence and Jira.

Following the successful rollout of these discovery tools, the company turned its attention to the development lifecycle. In early 2024, the focus shifted toward "Coding Agents," which provided developers with the ability to summarize code changes and suggest minor refactors. The current update marks the transition into "Agentic Loops"—autonomous workflows where an AI agent can scan a backlog, identify actionable tasks, write the necessary code, run tests, and propose a pull request (PR) for human review. This shift from "assistant" to "agent" represents a fundamental change in how software is architected and maintained.

Technical Deep Dive: Key Feature Enhancements

The newly introduced features are designed to provide both the flexibility needed by developers and the oversight required by engineering leadership.

1. Code Context and Teamwork Graph
Code Context allows AI agents to gain a holistic view of a multi-repository codebase. By leveraging the Teamwork Graph, these agents can now perform complex tasks such as vetting backlog ideas for architectural feasibility. Before a developer even writes a line of code, the agent can analyze whether the proposed feature conflicts with existing patterns, suggest implementation plans, and accelerate the root-cause discovery process during bug triage.

2. Agentic Loops and Jira Integration
The introduction of agentic loops in Jira is perhaps the most significant functional change. By continuously monitoring the backlog, these agents identify unassigned, well-defined work items. Once identified, the Jira Coding Agent takes over, executing the work, running automated tests, and submitting a PR directly back into the Jira interface. This creates a seamless, continuous feedback loop that reduces the administrative burden on software engineers.

3. Governance and Standards
To prevent "shadow AI" and ensure code quality, Atlassian has introduced "Standards" and "Agent Context Controls." Standards allow platform teams to define organizational coding conventions—such as naming patterns, security requirements, and library usage—which are then mapped to specific repositories. An AI-powered review agent then audits all incoming pull requests against these standards, flagging non-compliant code before it is ever merged. Meanwhile, Agent Context Controls provide granular oversight, allowing managers to define exactly which repositories an agent can access and what actions it is permitted to perform within a given space.

Measuring Success: DX for Agentic Development

A critical component of this rollout is the "DX for Agentic Development" framework. Engineering leaders have long been skeptical of AI adoption due to the difficulty of quantifying its impact. Atlassian is addressing this with a suite of metrics that track throughput, quality, and adoption.

The framework includes AI Code Insights, which provides a detailed breakdown of how much of a codebase was touched by AI versus human developers. Furthermore, it tracks "model-to-task fit," helping organizations understand which models perform best for specific types of work, such as backend refactoring versus frontend UI development. By mapping total AI investment to tangible engineering outputs, Atlassian is attempting to provide the data that CIOs and CTOs need to justify AI expenditures to their boards.

Broader Industry Implications

The move by Atlassian mirrors a broader trend in the software industry: the "professionalization" of AI agents. As firms like GitHub, GitLab, and Microsoft continue to refine their own AI assistants, the competitive advantage is shifting toward platforms that offer the best integration, governance, and measurement.

The implications for the developer experience (DX) are significant. Proponents argue that by automating the "drudgery" of software development—such as writing boilerplate code, updating documentation, and triaging minor bugs—AI agents will allow human engineers to focus on complex architectural decisions and high-level product strategy. However, skeptics caution that the reliance on AI-generated code could lead to a "black box" effect, where developers may struggle to debug code they did not personally write, potentially leading to long-term maintenance debt.

Atlassian’s emphasis on "accountable and measurable" AI suggests that the company is aiming to mitigate these risks. By requiring human review for PRs and keeping the "human-in-the-loop" as a central tenet of the agentic workflow, the firm is positioning its tools as a bridge between current manual practices and a more autonomous, AI-driven future.

Availability and Future Roadmap

The rollout of these features is staggered to ensure stability and enterprise readiness. Code Context is currently moving into an open beta for paid Atlassian customers, providing early adopters the chance to experiment with multi-repo awareness. More sophisticated features, including agentic loops, standards enforcement, and automated AI review, are currently available in private early access, with general availability expected in the coming months.

As these tools move toward general availability, the industry will be watching closely to see how effectively they scale. For large enterprises with thousands of repositories and complex compliance requirements, the ability to centralize AI governance in a platform like Jira could be a decisive factor in their choice of tooling. Atlassian’s commitment to providing an "agent-aware" SDLC is not just an incremental improvement to its software; it is a clear statement that the future of enterprise engineering will be defined by the successful, governed, and measurable integration of autonomous agents into the team’s collective intelligence.