Over three days in Amsterdam, Atlassian Team ’26 Europe delivered more than 200 announcements and one central idea that completely changes how we use Jira, Confluence and Jira Service Management: the era of “single-player” AI is over.
Until now, typical adoption involved one person typing alone in a chat whose context disappeared when it was closed. In our latest technical and strategic webinar—where we condensed the event’s essentials into 60 minutes—we explored how Atlassian is evolving from a set of productivity tools into a platform for collaborative human-agent work.
Watch the full session here:
Watch the webinar: A New Way of Working — Atlassian Team ’26 Europe Updates
1. The AMP protocol: agents with identity, a cursor and auditing
The major architectural building block unveiled at the event is AMP (Agentic Multiplayer Protocol), which enables agents to work within a team in real time with full traceability:
- Identity and ownership: Each agent is registered with its own profile, an assigned owner and granular permissions defined by the administrator.
- Real presence: You can
@mentionagents in Confluence or Jira, collaborate with them in Loom, or see their cursor editing a document live alongside their human teammates. - Unified context: They access the Teamwork Graph, which already connects more than 250 billion objects and supports over 80 native connectors, to understand projects, code and decisions rather than working blindly.
- Complete traceability: Every action is recorded in the audit log: who launched the agent, which data it accessed and what it changed.
2. Key platform and Rovo updates
Teamwork Graph, Atlassian Insights and MCP v2
Atlassian’s knowledge graph is extending its reach beyond its own tools:
- Rovo Code Search: Natural-language search across repositories and pull request history in Bitbucket and GitHub.
- Atlassian Insights (available in December 2026): Native connections to data lakes and warehouses such as Snowflake, BigQuery and Databricks, while respecting source permissions.
- MCP v2 server: Connects more than 220 external tools, including IDEs such as Cursor, Windsurf and VS Code, plus Figma, Slack and Zoom, reducing token consumption per task by up to 25%.
Rovo Work and Artifacts
- Rovo Work (Early Access): A new execution mode for long, multi-step tasks. You define the goal—for example, planning a sprint, drafting a post-mortem from closed tickets, or creating a 30-60-90-day onboarding plan. Rovo proposes a plan and, once you approve it, executes it in Jira and Confluence.
- Artifacts (Open Beta): A permanent space with a fixed URL for AI-generated output, such as interactive prototypes, calculators, dashboards and proposals. These can be embedded in Confluence, Jira or Slack with one click and respect space permissions.
3. Highlights from each Atlassian Collection
Atlassian structures its offering around its collections, integrating specialized agents into each workflow:
- Teamwork Collection: Highlights include Planner (Early Access), which turns an idea into a Confluence specification and breaks it down into Jira tasks; Loom AI Overlays; and the removal of the maximum user limit in Jira and Confluence Cloud.
- Service Collection (JSM): Introduces Solution Composer (Beta), which can configure request types, SLAs and portals from a description; automatic change risk analysis; Hardware Asset Management; and Proactive Service Management (EAP) to detect incidents before users report them.
- Software Collection: With Agent Sessions, developers work locally in Cursor, Windsurf or VS Code while keeping their Jira board synchronized through AMP. Other additions include an automated Code Reviewer, video PR reviews with Loom and AI impact metrics through DX AI Measurement. At Atlassian itself, 77% of pull requests are already generated by agents.
- Strategy Collection: Enables comparison of AI spending against budget by provider, team and priority (Open Beta), uses Rovo to infer team skills and reallocate capacity, and connects Focus and Funds to external assistants through MCP.
4. Governance, security and AI processing in the European Union
For European organizations and regulated environments, the webinar focuses on three critical advances in sovereignty and control:
- AI inference in the EU: LLMs now run on European servers, adding AI processing to data residency at rest and in transit, without transatlantic transfers and aligned with GDPR, ISO 42001 and the EU AI Act. Atlassian also plans to complete ENS and ISO 27017 certifications before the end of 2026.
- Non-Human Identity (NHI) inventory: Every agent, app and service account has its own identity, which can be managed alongside human users.
- Centralized control in Atlassian Guard: A single dashboard to audit which data each agent accesses, revoke access instantly and require human approval (human-in-the-loop) for sensitive actions such as deleting or publishing information.
5. Rovo credits: what changes on December 3, 2026
One of the session’s most practical topics is the Rovo credits consumption model. There is currently a grace period with no additional cost, but from December 3, 2026, overage billing will begin at USD 0.01 per credit, enabled by default unless a cap is configured.
Each organization has a single monthly pool, which resets every month and is calculated by adding the active users of each product:
| Product / Collection | Standard | Premium | Enterprise |
|---|---|---|---|
| Jira / Confluence / Loom | 25 credits/user | 70 credits/user | 150 credits/user |
| Teamwork Collection | 250 credits/user | 700 credits/user | 1,500 credits/user |
| Service Collection | 250 credits/user | 700 credits/user | 1,500 credits/user |
Practical example: An instance with 200 Jira + Confluence Premium users (14,000 credits) and 50 JSM Premium agents (35,000 credits) has a monthly pool of 49,000 credits. If it consumes 55,000 in one month, the 6,000-credit overage will cost an additional USD 60.
Autonomous agents and automations without human involvement are the biggest consumers. To optimize spending, we recommend migrating connectors to MCP v2, which saves up to 25% in tokens, reviewing unnecessary automations, and configuring alerts at 80% and 100% in Admin → Insights → Rovo credits.
6. Quabu’s recommended roadmap
To put these capabilities into practice without risks or budget overruns, we propose a three-phase plan:
- Phase 1 — Now (Governance and control): Audit current consumption in Insights, calculate your actual credit pool, configure spending caps before December 3, inventory MCP apps and review Atlassian Guard policies.
- Phase 2 — Q1 2027 (Controlled pilots): Select a small team and test a specific use case with Solution Composer, Artifacts or Rovo Work on a repetitive process.
- Phase 3 — Q2 2027 (Scaling): Extend agents to the rest of the organization with clear return metrics and optimized consumption.
Shall we review your instance together before December 3?
At Quabu, we help technical and business teams audit their Rovo credit consumption, configure agent governance in Atlassian Guard and launch their first AI pilots.
Email us at hello@quabusolutions.com or contact us through quabusolutions.com to schedule a review session with our consultants.
Translated from the original Quabu article.
