Atlassian co-founder and CEO Mike Cannon-Brookes and his team unveiled a “multi-player” agentic AI system at the supplier’s customer conference in Amsterdam.
In a press statement, he said: “The best work has never been a solo act, and now that’s true of AI too. The companies that pull ahead will be the ones that get their people and agents working together in the flow, out in the open, as one team. AMP [Agentic Multiplayer Protocol] is how Atlassian is making that real for teams.”
In a press briefing on the eve of the opening keynote, which ended abruptly around the 90-minute mark with a power outage, he said: “The agent experience is largely single player today. It is largely on my computer, I talk to an agent, maybe it’s running in the cloud or running locally, and it’s me back to the agent and I get some output and I take the output somewhere else. It’s not a very multiplayer and collaborative experience.”
He presented the Atlassian AMP as way of avoiding lock in to one particular frontier large language model (LLM): “Most customers, well north of 75% of our customers, are using multiple large-scale vendors. They’re picking multiple foundation model vendors, often with Microsoft in the mix, often with Google in the mix. Most big enterprises aren’t saying, ‘I’m going to pick A’.”
Sherif Mansour, head of AI at Atlassian, told Computer Weekly in a briefing at the event that the supplier sees the significance of AMP as moving customers more towards team working than working as an individual.
“If you look at the market today, most people, when they think of an AI experience, think of a chat prompt with an agent. You see this all the time: ‘Look what I can do with Claude’, let’s say. That’s awesome for individual productivity, but what if you are working on a document with another teammate and an agent?
“We build a lot of apps for collaborative teamwork, usually technology teams and business teams, bridging those two worlds together. So, we typically work not on single player products, but multiplayer products. And now you can bring in multiple specialist agents to do that.
“We allow other app vendors to be interoperable with the Atlassian platform and the Atlassian agents to be interoperable with their platforms with our AMP. It’s huge part of AMP to let customers pull in the agents they want. It doesn’t have to be ours.”
Atlassian’s own AI platform is Rovov, which launched in late 2025.
According to the supplier, as AI shifts to “multiplayer mode”, agents operate in the flow of work, collaborating alongside teammates in the open. This is said to keep an agent’s work visible to its whole team, preserving context.
AMP allows agents to participate wherever work happens, on the Atlassian platform through @mentions in Confluence, Jira comment threads or Loom video briefs, while the Atlassian MCP [model context protocol] server can pull in “context” from third-party tools such as Figma or IDEs.
Each agent under AMP is said to have a clear owner and distinct profile. The agents appear in real-time presence bars and cursors alongside human teammates.
The agents are also said to be grounded in the supplier’s Teamwork Graph, which connects more than 250 billion objects and relationships. The graph now reads source code right down to the function, symbol and class level, with a new Code Search app.
Atlassian contends that as AI adoption grows, work increasingly happens where leaders can’t see it, such as in terminals, local sessions and third-party bots.
The supplier said it makes that work visible, traceable and governable, and “agent sessions” surfaces cloud and local agent work in Jira and the Teamwork Graph, so context carries forward instead of getting lost in terminals.
In a press briefing on the eve of the main keynote, Dave Meyer, head of product at Atlassian, said: “The models are getting better and better … becoming super intelligent and yet we are not seeing super revenue at every company that’s building software.
“We’re building more software, but the coordination between teams is becoming harder. We’re seeing more quality problems, more review problems, more rework because we have to re-implement and re-architect systems multiple times because it wasn’t fully built or fully planned out in our enterprise architecture to begin with.
“Those are the large-scale problems, but I think we’re only scratching the surface of the more low-level knowledge and context loss that’s happening.”
