Atlassian introduces AMP for collaboration between people and agents

Atlassian introduced AMP, the Agentic Multiplayer Protocol, as a foundation for collaboration between people and agents. The company presents it as a move away from AI use by individuals working alone toward agents working alongside human teammates. AMP is intended to give agents context, visibility and permissions to contribute across Atlassian and other tools. The protocol is organized into three connected layers: in-the-flow collaboration, identity and shared presence, and context and governance.

The first layer places agent interactions in the tools and workflows where teams work. Atlassian says people can brief agents with Loom videos, mention them in Confluence documents, or discuss approaches in Jira comment threads. Agents can also be connected to Jira workflows for triage and routing, while MCP connections can bring tools such as Figma or an IDE into the workflow. In another described interaction, a person’s voice, cursor and selections become a brief with acceptance criteria; the agent then works and reports back for review and refinement. These examples combine ways to direct agent work with opportunities for people to inspect it.

The identity and shared-presence layer is meant to make agent activity visible and attributable. Atlassian says every agent has an owner and profile, with real-time attribution for its work. When people edit a document or view a whiteboard, they can see an agent’s cursor alongside those of human teammates. The company also describes showing the changes agents make so teams can trace contributions and review how content evolved. Together, these features are presented as ways for teams to understand who is responsible for an agent and how it is contributing.

The third layer focuses on context and governance. Atlassian says agents draw institutional context from the Teamwork Graph, which connects information across SaaS tools. Permissions can vary by role: the company contrasts an HR agent that inherits the prompting person’s access with a Helpdesk agent using a dedicated service account. It also describes an audit showing who triggered an agent and what data it touched, along with human checkpoints. For AI builders, AMP’s design treats workflow integration, agent identity, scoped access and review as parts of team collaboration, alongside the agent’s task capabilities. Atlassian says AMP begins rollout today.