Blog - Isos Technology

Rovo & MCP: A Partner’s Guardedly Optimistic View

Written by Michael March | Jun 24, 2025

Quick take

After a shaky first lap, Atlassian has started to put real horsepower behind Rovo. The product that felt like a science‑fair demo a year ago is now bundled into paid tiers, backed by 50+ connectors, and flanked by a growing agent studio. That’s progress worth cheering—just maybe keep the champagne on ice until we see sustained usage outside the Atlassian bubble. Meanwhile, the Remote MCP Server and the wider Model Context Protocol ecosystem give partners like us new pipes to expose Jira/Confluence data to other AI stacks. Together they form a classic hub‑and‑spoke pattern: Rovo pulls everything into Atlassian; MCP pushes Atlassian data out. Both stories are trending positive, but neither is fully written.

Where we’ve been – a choppy first year

  • Launch hype, limited bite. Rovo debuted at Team ’24 with promises of universal search and AI teammates.

  • Beta frustration. Early testers called the experience “impressive… but non‑deterministic” and flagged missing guard‑rails in follow‑up prompts.

  • Agent business‑risk threads. Devs worried about model drift and lack of versioning control in the public forums.

  • Adoption stayed niche. By late 2024 Atlassian itself admitted that most usage was still inside internal dog‑food tenants, even as closed‑beta customers reported 1–2 hrs per week saved on average.

Team ’25: momentum shift

  • “Rovo for everyone.” In April 2025 Atlassian announced that Rovo will become included for Premium and Enterprise customers .

  • Connector fire‑hose. The blog post highlighted 50 out‑of‑the‑box connectors and a promise that partners can build their own via Rovo Studio.

  • Investor jitters. Analysts questioned the free‑tier move after Atlassian posted a quarterly loss and saw its share price drop 16 %.

  • Reality check. Confusion still crops up— one January 2025 email even declared “Rovo is being retired,” underscoring comms growing pains.

Why partners can feel cautiously upbeat

  • Studio & CLI = developer gravity. The new Rovo Studio and the just‑launched Dev CLI let builders craft agents and wire in their own MCP servers—showing Atlassian is serious about extensibility.

  • Usage signals. Atlassian claims over 1.5 M monthly active Rovo users and SWE‑bench‑leading scores for the Dev Agent, hinting at product‑market fit beyond marketing decks.

  • Community tone improving. Forum sentiment has shifted noticeably: a May 2025 prompt‑engineering thread on the Atlassian Intelligence space racked up 200 up‑votes in two days; the latest Rovo AMA saw more “how do we get even more value?” questions than “why doesn’t it work?” laments. Support inquires seen to have swung from break‑fix to best‑practice. That’s a small sample size—but directionally positive.

  • Org‑wide switch‑on. Because Rovo lives inside Atlassian Cloud, admins can enable it company‑wide with a single toggle—making enterprise rollout the easy path. MCP, by contrast, expects each user or team to configure their own client, so adoption is piecemeal and enablement‑heavy.

Enter MCP – spokes for the wider AI wheel

  • Remote MCP Server . As of May 2025 Jira & Confluence can now be queried from Anthropic Claude (and other MCP-speaking clients) via Atlassian‑hosted endpoints, complete with OAuth passthrough and Teamwork Graph permissions.

  • Industry rally. MCP has become a poster‑child for open AI plumbing, drawing comparisons to “USB‑C for LLMs” and sparking think‑pieces about protocol wars.

  • Rovo already nibbling. The Dev CLI docs explicitly show how to point an agent at your own MCP server, a hint that the two paths are converging.

  • Security spotlight. MCP is barely a year old (introduced Nov 2024) so the red‑team crowd is still mapping its attack surface. Recent write‑ups from Microsoft Defender and Palo Alto Networks flag risks around misconfigured servers, unsecured tokens, and evolving auth best‑practices—reassuring that vigilance, not blind trust, is the order of the day.

Hub vs. Spoke Recap

Pattern

 

Primary Flow

 

Best When

 

Watch‑outs

 

Rovo

External data → Atlassian

Teams live in Jira/Confluence & want turnkey AI search/chat/agents

Connector quality; opaque model capabilities

Remote MCP

Atlassian data → External AI tools

Org already uses multi‑LLM stacks, RAG pipelines, IDE agents

Spec still evolving; currently Jira & Confluence only; requires per‑user client configuration; security best‑practices still maturing

Open questions for H2 2025

  1. Connector depth vs. breadth: 50+ logos is great; how many deliver action APIs vs. just search index?

  2. Pricing cliff: Will “included” Rovo stay free in FY 26, and what metering will come with agents?

  3. Security & governance: How will Atlassian surface audit trails for agent actions, especially when routing through MCP?

  4. Ecosystem split risk: If MCP adoption accelerates, will Atlassian migrate Rovo connectors to the same protocol or run two parallel stacks?

Bottom line

Rovo has gone from proof‑of‑concept to promising platform in roughly twelve months, thanks to a pricing flip, a connector surge, and the first taste of developer tooling. The Remote MCP Server complements that story by letting external LLMs tap Atlassian data without reinventing APIs. For partners, the mix spells genuine opportunity—but with enough unanswered questions to justify cautious optimism rather than hype. Keep building, keep asking hard questions, and keep your options open.