Atlassian launches AI agents and visual tools in Confluence to transform documentation workflows

Atlassian launches visual AI tools and third-party agents in Confluence | TechCrunch

Software giant Atlassian announced new AI tools and agents on Wednesday with a focus on turning data into visual assets and applications. This includes the rollout of the visual tool Remix in open beta. Remix allows enterprises to turn the data and information stored in Atlassian's content collaboration software Confluence into assets including charts and graphics. Remix will recommend which visual format makes the most sense for the data or information at hand and create these visual assets without requiring the users to open another application or software. The company also announced three new third-party agents that run within Confluence using model context protocols (MCPs). One agent connects Confluence users to the vibe coding darling Lovable to turn product ideas and data into working prototypes. Another agent connects to app builder software Replit and allows users to convert technical documents into starter apps. The third agent works with AI presentation builder Gamma to build slides and other presentation materials. "With Remix and agents in Confluence, a single page becomes the starting point for whatever comes next: a clear story for leaders, a prototype for builders, or a walkthrough for customers, all from the same source of truth," Sanchan Saxena, senior vice president of teamwork collaboration at Atlassian, wrote in a blog post announcing the features. "When you remove that friction, teams do more than manage documents; they create the next generation of products and experiences." The new tools are the latest in Atlassian's push to incorporate AI agents and tools directly into the apps workers are already using, as opposed to launching new software platforms. In February, the company added AI agents to its product management software Jira. This follows a trend across the industry of companies looking to embed AI tools and agents directly into existing workflows as opposed to launching separate AI-powered software. While Salesforce was one of the first enterprises to launch a separate AI agent management platform, Agentforce, in 2024, its since released many of its AI innovations through existing software like its recent upgrade that turned messaging service Slack's chatbot into an AI agent. OpenAI is also leaning into this movement through its recent Frontier Alliances initiative. OpenAI partnered with four major consultant firms to task consultants with embedding OpenAI's tech into their clients' existing tech stacks and workflows as opposed to just selling them ChatGPT Enterprise subscriptions. "Technology should fade into the background and let people focus on their best work," Saxena wrote in the company's blog post.

Atlassian gussies up Confluence for the AI era

Atlassian is modernizing Confluence for the AI era, testing tools and agentic capabilities that give users the chance to turn their written notes into graphics and their ideas into software applications. Each product is in its early stages and is being tested with a small group of customers before a wider deployment, an Atlassian spokesperson said via email. It is part of a strategy that means IT teams have to deal with fewer bespoke requests and security risks, since in each case the AI is being brought to the data inside Confluence. For employees, Remix with Rovo - Rovo is Atlassian's AI assistant - gives them several ways to present the data that they have housed inside Confluence, the company's team workspace for managing projects. "They can turn static docs, tables, or unstructured data into the format or workflow they need, tailored for the right audience or altitude -- without leaving Confluence or opening a ticket," the spokesperson told The Register. "Our data found that Confluence pages with visual elements are nearly two-times as likely to be read by a wider audience compared to pages without." The product appears to work like Google's Notebook LM, which can also manipulate several file types to present data as a podcast, graphics, or a slide deck. However, Atlassian said Notebook LM has limits. "Notebook LM is great for multi-modal transformation, but it works in a vacuum. The key difference is workspace-native context," the spokesperson wrote. "Remix works within the pages, permissions, and structures teams already use -- meeting notes, PRDs, runbooks -- not a separate environment you copy/paste back from. With Notebook LM, users run the risk of creating an isolated artifact - with Remix, the output is always connected to the source content." Notebook LM, they said, is a single-player experience, while Remix outputs live inside Confluence where teams already work together. Comments, mentions, and real-time editing are native, not bolted on. "The content stays multiplayer from the start, organized and findable by the entire team," Atlassian's spokesperson said. On the agentic front, Atlassian said it has partnered with Lovable, Replit, and Gamma while promising more agent providers to come to turn data inside Confluence into software applications. Confluence already allows users to create artifacts such as product requirements documents that can link directly to work tracked in Jira, its project management product. These new partner agents simply compress the process of taking a product requirements document, prototype, or strategy note and turning it into an app, prototype, or presentation using AI. "It's not about turning Confluence into an app factory. It's about letting teams transform knowledge into whatever format it needs to be - with the source knowledge and the resulting experience connected and governed in one place," an Atlassian spokesperson said. The agents have no independent ability to carry out tasks without the user's permission. "They don't silently deploy apps or make architectural decisions on their own," Atlassian's spokesperson said. "The user initiates. They might say 'turn this into an app' or ask what's possible - and the agent suggests options and scaffolds a starting point. Users review and confirm outputs, but the experience goes further: teams can set up automations where partner agents proactively act on a schedule or trigger, without manual prompting." In terms of guardrails, Atlassian said its agents operate within Confluence's existing access controls, meaning that if a user can't see a page, the agent can't either. Users must also review and confirm outputs before anything is published or deployed. "Teams that want to go further can configure agents to act proactively: triggered by a schedule, an event, or a workflow condition. The level of autonomy is yours to dial." In March, the company announced it would lay off about 10 percent of its staff, cutting about 1,600 jobs to fund AI initiatives. Late last year, the Australia-based company migrated more than 3,000 Jira and Confluence instances to AWS Graviton processors, with Graviton 4 handling user-facing tasks, resulting in roughly 10 percent savings, lower latency, and better customer response times. ®

Atlassian brings AI visual tools and partner agents to Confluence, 1 month after cutting 1,600 jobs

In short: Atlassian is rolling out Remix, a visual AI tool in open beta that transforms Confluence pages into charts, infographics, and scorecards without requiring users to open another application, alongside three partner agents built on the Model Context Protocol that will carry Confluence content directly into Lovable, Replit, and Gamma from April 13. The announcement arrives less than a month after Atlassian cut 1,600 jobs explicitly to fund AI investment. Knowledge management software has a presentation problem. Teams invest enormous effort documenting decisions, specifications, and meeting outcomes in Confluence, and then spend comparable effort manually reformatting that same content into the charts, prototypes, and presentations that different audiences actually need. Atlassian is attempting to close that gap with two interconnected announcements on Wednesday: a visual generation tool called Remix that keeps output tethered to its source, and a set of pre-built agents that hand Confluence content directly to partner applications. Remix, now in open beta, allows teams to highlight any content on a Confluence page, a paragraph, a table, or an entire document, and instruct the tool to generate a visual from it. At launch, supported output formats include data visualisations, infographics, scorecards, and charts, with Atlassian stating additional formats will be added over time. The resulting visual is layered on top of the original content and linked to the source, meaning it updates as the underlying page changes and does not require a separate export or file management workflow. The intelligence guiding Remix's format recommendations comes from the Teamwork Graph, Atlassian's unified data layer built from more than 100 billion data points across Jira, Confluence, and connected enterprise tools. Rather than asking users to choose a format manually, Remix uses that graph to surface the visual type most likely to be useful given the content's structure and the organisation's usage patterns, a quarterly roadmap page, for instance, might prompt a scorecard; a dataset might prompt a chart. Sanchan Saxena, Atlassian's senior vice president of product for the Teamwork Collection, framed the tool as an attempt to make the platform recede: "With Remix and agents in Confluence, a single page becomes the starting point for whatever comes next: a clear story for leaders, a prototype for builders, or a walkthrough for customers, all from the same source of truth." Where Remix keeps output inside Confluence, the partner agents announced alongside it are designed to move content out of Confluence and into specialist tools without any manual copying or custom integration work. Three agents are launching on April 13: Lovable, which converts a product specification into a working user interface prototype; Replit, which turns a technical document into a starter application that an engineer can fork and build upon; and Gamma, which transforms meeting notes or a status page into a polished presentation. Each agent is invoked directly from a Confluence page through Rovo Chat. When triggered, the agent reads the page's content and metadata, including authorship, project association, and decision context, and carries all of it into the partner tool without requiring the user to manually reconstruct that context on the other side. The artifact produced, a prototype in Lovable, a codebase in Replit, a deck in Gamma -- links back to the source page it came from, preserving the chain of reference between documentation and output. For administrators, setup requires no custom scripting. Enabling a partner's Model Context Protocol server in Atlassian Administration takes a matter of minutes, after which the agent appears in the team's Rovo directory, pre-configured by the partner and inheriting the workspace's existing permissions and context. The technical foundation for the partner agents is the Model Context Protocol, the open standard that has rapidly become the connective tissue of the agentic software ecosystem. Atlassian's choice to build on MCP rather than a proprietary integration layer is a deliberate strategic signal: any partner can build an agent that works with Confluence content without waiting for Atlassian to construct a bespoke connection. The protocol is open and the server documented, meaning the barrier to joining the ecosystem is technical competence rather than a bilateral commercial agreement with Atlassian. The three launch partners span different use cases by design. Replit, which also features as a launch partner in Anthropic's recently announced enterprise software marketplace, represents the developer workflow; Lovable the product design and prototyping workflow; Gamma the executive communication workflow. Together they cover the three primary audiences for whom Confluence documentation most consistently needs to be reformatted before it becomes actionable. Atlassian cut 1,600 people from its payroll in March, approximately 10% of its global workforce, with chief executive Mike Cannon-Brookes stating that the savings would be redirected into AI investment and enterprise sales. The company simultaneously replaced its chief technology officer, splitting the role between two executives: Taroon Mandhana as CTO Teamwork and Vikram Rao as CTO Enterprise and Chief Trust Officer. Remix and the partner agents are, in effect, the first significant product announcement since that restructuring, and a direct demonstration of what that investment is intended to produce. The competitive pressure is real. Microsoft's own terms of service now describe Copilot as for entertainment purposes, a characterisation that surfaced as Copilot's accuracy Net Promoter Score dropped to -24.1 by September 2025, with nearly half of lapsed users citing distrust of its answers as the primary reason they stopped using it. Atlassian's approach, embedding AI into workflows that already contain verified organisational context rather than asking users to interact with a general-purpose chat interface, is a direct response to that failure mode. Rovo has reached five million monthly active users, according to Atlassian's own reporting, suggesting the positioning is landing with enterprise teams even as the company's share price has reflected broader investor anxiety about conventional SaaS tools in an agentic AI era. Workplace AI adoption surged through 2025, with tools like Microsoft 365 Copilot becoming common fixtures on office desktops, but the consensus among enterprise technology teams shifted from enthusiasm to scepticism as accuracy issues and context limitations became apparent in production. Atlassian's bet with Remix and the MCP-based agents is that the solution is not better general-purpose AI but AI that is anchored in the specific knowledge a team has already produced, and that the role of a platform like Confluence is less to store that knowledge than to make it continuously available in whatever form the work requires next.