Atlassian rolls out tools to trace AI-generated code
Atlassian introduced Agentic Multiplayer Protocol and related tools to help enterprises track whether code came from humans or AI agents. The offering includes Teamwork Graph for indexing source code and showing attribution across tools such as Bitbucket, GitHub, Claude, Codex, Figma, and Rovo, plus a Rovo Work mode for multi-step tasks with human approval. Analysts welcome greater visibility but question whether the product provides enough useful context.
Atlassian's AMP aims to define how people and agents work together, with identity, limited authority, shared context, tasks, and reviewable results. Teamwork Graph indexes code down to functions, symbols, and classes, enabling searches across Bitbucket and GitHub without cloning repositories, while showing version history and attribution across Claude, Codex, Figma, and Rovo.
Rovo Work can carry out long, multi-step jobs across Jira, Confluence, and connected tools, requiring human approval and running in a secure sandbox. Atlassian also introduced EU-only LLM processing and promised non-human identity controls soon. Analysts say separating human and AI contributions is difficult, partly because Git attribution often relies on editable commit messages.
If such provenance tools work, software teams, auditors, and customers could gain clearer records of how code was produced, potentially improving accountability and review. Developers may face new expectations to document AI assistance, while enterprises might tailor oversight by risk. Yet incomplete or noisy attribution could create false confidence or extra bureaucracy. The broader effect may depend on whether organizations use these signals to support judgment rather than replace it.