# Atlassian Launches AMP for Human-Agent Teamwork

Published 2026-10-07 · AI.info Team · AI.info

> Atlassian announced the Agentic Multiplayer Protocol, or AMP, at Team ’26 Europe in Amsterdam on October 6, 2026. The framework connects agents to Atlassian’s Teamwork Graph and adds shared work features, review steps and controls for agent access.

[Atlassian](https://www.atlassian.com/blog/company-news/team26-europe-ai-platform) announced its Agentic Multiplayer Protocol, or AMP, on October 6 at Team ’26 Europe in Amsterdam, pitching it as a way for people and software agents to work in shared company workflows. AMP gives agents an identity, limited authority, shared context and tasks, and results that people can review. The announcement combines that framework with new features across Jira, Confluence, Rovo and Atlassian’s connected tools.

## AMP puts agents in the team’s work stream

Atlassian describes AMP as the foundation for agents to participate in collaborative work rather than operate only in separate chat sessions. An agent could take part in a Jira comment thread, receive a brief through Loom or Confluence, or handle a step in a Jira workflow. Agents can draw on the Teamwork Graph, which connects work, people, knowledge, code and other business context.

The graph contains more than 250 billion connections, and people and agents already collaborate more than 10 million times a month across Atlassian’s platform. Those are company figures, not independently verified measures of customer outcomes. The company frames AMP as a way to make that activity more visible and connected, with agents’ actions linked to shared work that teammates can inspect.

> “Atlassian provides the context we need to make AI work for us. We’ve proven that a good model powered by Atlassian’s organizational context will always outperform the best model in the world working in a vacuum.”
>
> [Jose Ramirez, Executive Director, Enterprise Agility & Delivery, Kaplan North America](https://www.atlassian.com/blog/company-news/team26-europe-ai-platform)

## Rovo Work and Artifacts carry tasks beyond chat

One of the announced features, Rovo Work, is a mode in Rovo Chat for multi-step tasks across Jira, Confluence and connected tools. A user sets a goal, reviews and adjusts the proposed plan, then approves execution. A task can run for hours in an administrator-governed secure sandbox, with human review built into the process.

The new Artifacts app gives AI-generated plans, interactive models and HTML views a persistent URL. Artifacts from Atlassian’s own tools or other AI products can be embedded in Confluence pages, Jira work items and Slack channels, subject to enterprise permissions, and indexed into the Teamwork Graph. The company’s rebuilt Model Context Protocol offering also expands from dozens of tools to more than 200 across Atlassian and other services.

## Identity and access controls are part of the pitch

AMP assigns agents identities and lets administrators scope what they can access. Agent accounts and an inventory of non-human identities are coming, giving administrators a way to manage agents, apps and service accounts and revoke access. The company also points to Atlassian Guard controls for discovering and classifying sensitive data, scanning content and attachments, and setting guardrails for Rovo Chat and connected tools.

Atlassian announced local EU AI inference, which it describes as restricting model processing to models hosted within the European Union, along with compliance support for standards including the EU AI Act and ISO 42001 for Rovo. The announcement does not give a single availability date for every feature: it says AMP begins rolling out, while the account and identity controls are described as coming soon. That distinction matters for organizations assessing which governance capabilities they can use now.

## A shared protocol, with rollout details still uneven

The announcement also adds Rovo Code Search for searching source code across Bitbucket and [GitHub](https://ai.info/companies/github) in plain language, structured data from lakes and warehouses through Atlassian Insights, and agent sessions in Jira that link local and cloud agent activity to work items. Atlassian’s connector library now exceeds 80 integrations, including additions such as Zoom, Gong, [Microsoft](https://ai.info/companies/microsoft) Entra ID and [Google](https://ai.info/companies/google) Identity. Together, the releases tie AMP to a broader effort to make company information and agent activity available across the software teams already use.

AMP is a collection of experiences, technologies and patterns for a shared human-agent operating model, in Atlassian’s words. Its practical test will be whether the promised shared context, traceable actions and permission controls work consistently across products and external tools—not just whether an agent can complete a task. For now, AMP is live across its platform, the company says, while several of the controls and experiences remain in rollout or are scheduled for later.

## Sources

1. Original source · Blog post [Inside Atlassian](https://www.atlassian.com/blog/company-news/team26-europe-ai-platform) atlassian.com

## Companies in this story

- [Atlassian](https://ai.info/companies/atlassian)

## Related stories

- [Google Introduces Gemini as a Universal Agent for Work](https://ai.info/news/google-gemini-universal-agent-work)
- [Cohere Adds Agent Memory and Spending Caps to North 2](https://ai.info/news/cohere-north-2-agent-memory-spending-controls)
- [Satya Nadella Calls for a Human-Controlled AI “Emergency Brake”](https://ai.info/news/satya-nadella-ai-emergency-brake)
- [OpenAI Has Notified Over 100 Organizations of Agent Activity](https://ai.info/news/openai-notifies-100-organizations-rogue-agent-activity)
- [Workday to Acquire C1.ai to Govern AI-Agent Access](https://ai.info/news/workday-to-acquire-c1-ai-agent-access)

## Explore

[More articles](https://ai.info/news)

---

Canonical: https://ai.info/news/atlassian-amp-human-agent-teamwork
