Atlassian rolled out a set of AI-native development features into Jira on July 15, and the headline number is the kind that’s actually worth checking: agents given access to Atlassian’s “Teamwork Graph” context layer produced 44% more accurate results while using 48% fewer tokens than agents working without it. That’s not a vague productivity claim, it’s a specific before-and-after comparison, and it points at something real: most AI coding agents fail not because the model is weak, but because they’re missing context about the actual codebase, the actual ticket history, the actual team conventions.
The Teamwork Graph is Atlassian’s attempt to fix that by connecting work items, code, goals, and team knowledge into one layer that agents can query. The most concrete result is the new Jira Coding Agent, which turns a work item directly into a ready-to-review pull request without anyone setting up a local dev environment first. You can also now assign a ticket straight to Claude, Cursor, or GitHub Copilot from inside Jira itself, rather than copying context between five different tabs by hand.
The bigger strategic move here is that Atlassian isn’t trying to build its own coding model to compete with Anthropic or OpenAI. It’s positioning Jira as the context layer that makes whichever model you already use work better, which is a smarter bet than trying to out-model companies with far more compute. Most of this is available to existing paid Jira Cloud customers at no extra cost, with separate tools for tracking token spend per pull request, since “the AI wrote code for free” and “the AI wrote code using a lot of paid tokens” are two very different budget lines.
Whether this actually changes how engineering teams work depends on something Atlassian can’t control: whether teams keep their Jira tickets and documentation clean enough for a context graph to be useful in the first place. A coding agent with perfect context and a messy backlog underneath it is still going to produce mediocre pull requests.
Atlassian isn’t alone in betting that the winning move is making existing agents smarter rather than shipping a new one. HubSpot’s Agent Hub makes almost the identical bet for sales and marketing teams: less “here’s a new AI agent,” more “here’s a single place to see and manage the ones you already have.” The pattern across the industry right now points toward context and visibility mattering more than raw model capability. Full details on the Teamwork Graph rollout are worth a look if you manage a Jira instance directly.