Agentic attribution in Confluence

When you use agents to do work across Confluence, the outputs of that work — for example, creating content, viewing or editing existing content, or adding a comment — are attributed to you and the agent. This applies to all types of agents, including Rovo, Atlassian Studio agents, Forge agents, and MCP-registered agentic systems outside of Atlassian that are used with Confluence.

Working with agents in Confluence

There are many ways users can invoke agents:

  • Mentioning agents directly in Confluence product surfaces.

  • Invoking Rovo or Studio agents through Rovo chat to view, edit, and create content.

  • Using agentic features that are powered by Rovo.

  • Prompting agents such as Claude, Codex, and ChatGPT to do work in Confluence over MCP or the Teamwork Graph CLI.

How agentic work is attributed

Agent attribution brings agent and teammate activity together, helping you see what agents contributed and who they acted on behalf of, when applicable.

There are two types of agent attribution:

  • Dual attribution: When an agent performs work on behalf of a user, Confluence displays both names as [Agent] with [Human].

  • Sole attribution: When an agent performs work independently (such as through an automation) or is configured to act as its own entity, the UI displays only the agent’s name.

Where in Confluence agentic work output is surfaced

  • Contributors and version history: When an agent edits or updates content, they are listed as a contributor.

  • Analytics: When an agent views a content item, they appear in the analytics reports (pages, live docs, and blogs only).

  • Comments: When an agent makes a comment, they are listed as the author.

  • Notifications: When an agent mentions you or updates a page you follow, the notification states which agent performed the action and, for dual attribution, who invoked it.

  • Avatar visibility: When agents are viewing, editing, or commenting on content, their avatars appear in the content’s live presence avatar group.

  • Audit logs: If Confluence records an audit log for an action that an agent took part in, the log will carry the agent attribution as well.

How work is attributed to agents

Confluence leverages Atlassian-level agent identities to surface the activity and work outputs that agents have done across the app. Depending on the type of agent or agentic system, there are differences in how the agent can be identified and represented across product surfaces.

This is not a compliance or audit tool. If you’re looking for compliance tooling, explore Atlassian Guard.

Type of agent or agentic system

How the agent is identified

What attribution looks like across Confluence

Rovo

Rovo is a system user with an established agent user identity

Rovo with Annie Claire

The attribution of a user and Rovo together.

Atlassian Studio agents

When a new Studio agent is created, a system user for that agent is created

Plan Builder with Annie Claire

A studio agent attribution with a user.

MCP-registered agentic systems using MCP v2

When MCP v2 agentic traffic is registered, it declares who it is — Claude, Cursor, ChatGPT — and an agent user identity is generated

Claude with Annie Claire

A Claude agent and a user attribution.

MCP-registered agentic systems using MCP v1

When MCP v1 agentic traffic is registered, it all appears as a single user, regardless of the underlying agent harness. A single agent user identity is generated for all MCP v1 agentic traffic.

External agent with Annie Claire

An external agent attribution with a user.

Agentic systems using Teamwork graph CLI

When Teamwork graph CLI agentic traffic is registered, it all appears as a single user, regardless of the underlying agent harness. A single agent user identity is generated for all Teamwork graph CLI agentic traffic

External agent with Annie Claire

An external agent attribution with a user.

Dive deeper into Atlassian Rovo MCP Server and Teamwork Graph CLI.

Still need help?

The Atlassian Community is here for you.