Itdaily - Atlassian expands Teamwork Graph for AI agents and developers

Atlassian expands Teamwork Graph for AI agents and developers

Banner reading 'Teamwork Graph' surrounded by profile photos and status labels like 'On track' and 'At risk'.

Atlassian is implementing several expansions to Teamwork Graph, the context layer behind its AI approach. The company is making the solution more widely available, adding new tools for developers and knowledge workers, and now allowing customers to connect their own data sources.

Atlassian expands Teamwork Graph for AI agents, developers, and other users. The company is announcing these improvements during its Team ’24 conference. The innovations focus on broader access to the context layer, which connects people, goals, code, and content across Atlassian services and linked SaaS apps.

Context in more tools

First and foremost, Teamwork Graph is becoming available in more environments, including browsers, mobile apps, and the terminal. Atlassian aims to ensure that AI tools no longer use isolated pieces of information but work with the same context that teams are already building in their existing workflows.

According to Atlassian, this context layer should make the difference between AI that guesses and AI that knows where a decision comes from, who owns the work, and what connections exist with other projects. This is relevant for organizations that want to deploy AI agents without them producing hallucinated answers. Atlassian states that AI using Teamwork Graph data in this way is 44% more accurate and requires 48% fewer tokens to process a given query.

New tools for developers and assistants

For developers, Atlassian is introducing the Teamwork Graph CLI in open beta. With over three hundred commands, technical users can query the Graph directly from the terminal and pass that data to AI tools, agents, or custom workflows. The CLI is not just for reading; developers can also modify relationships and work items.

Additionally, Atlassian is bringing Teamwork Graph developer tools to the Rovo MCP Server, also in open beta. This allows AI assistants that work with MCP, such as external assistants, to not only retrieve data but also perform actions. This should be useful for incident response, smarter notifications, and finding a single source of truth across Jira and Confluence.