Windsurf Upgraded and Renamed to Devin Desktop

On June 2, Windsurf users opened their editor and received an OTA update. After installation, it is now called Devin Desktop.

From Codeium to Windsurf to Devin Desktop, this product has been renamed three times in three years. But this is not just a logo change: the default interface has shifted from "editor-first" to an Agent Command Center, a Kanban-style board that unifies management of all your local and cloud agent sessions.

The editor has stepped back, and the agent has stepped forward.

Three layers of change in this update

Brand layer: Windsurf is retired, and the product is merged into the Devin system. Plans, pricing, extensions, and shortcuts are all retained, and the official statement confirms no price changes. Existing users experience basically frictionless migration.

Engine layer: Local agent upgrade. The original Cascade is replaced by Devin Local, rewritten from scratch in Rust. The official claim is token efficiency improved by up to 30%, with native support for subagents. The old Cascade is retained until July 1, 2026, after which it will be taken offline. Users still using Cascade workflows need to complete migration this month.

Protocol layer: Built-in Agent Client Protocol (ACP) support. This is the point most worth expanding on.

ACP: Decoupling editor and agent

The problem ACP solves can be stated in one sentence: let any agent run in any editor.

In Devin Desktop, you can directly run third-party agents such as Codex, Claude Agent, and OpenCode, sharing the same UI and context. As of June, ACP has been adopted by JetBrains, Google, GitHub, and others, with over 25 connected agents.

The reference point for this is LSP (Language Server Protocol). Before LSP, each editor had to write support for each language separately, an N×M adaptation disaster; after LSP, language services written once could be used everywhere. ACP is doing the same for agents: editors are no longer bound to their own agents, and agents are no longer bound to their own editors.

For Cognition, this move is somewhat counter-intuitive: they spent a lot of money to acquire an IDE, yet actively integrate competitors' agents. But the logic holds: rather than betting users will only use Devin, bet that users need a place to manage all agents. Selling a workbench is better than selling shovels.

Does "IDE as Agent Manager" hold up?

I used the Agent Command Center perspective for a few days to think about this question. The conclusion is: the direction is likely correct, but the pace might be too fast.

The direction is correct because multi-agent parallelism is already the real daily routine for heavy users. Opening three or five tasks simultaneously—one fixing bugs, one writing tests, one refactoring—this workflow cannot be supported by traditional editor tab management; a Kanban board is a more reasonable form.

The pace is too fast because most developers are still at the "one conversation at a time" usage stage. Setting Kanban as the default view is a cognitive tax for this portion of users. Fortunately, the editor view is still there, and one shortcut can switch back.

The judgment worth remembering is: when JetBrains, Google, GitHub, and Cognition all bet on the same protocol, this protocol is likely to win. Looking back in 2027, "which agent is built into the editor" might become as outdated a question as "which language the editor supports".

Reference Source: Devin Official Blog / Official FAQ.

This page provides an editorial summary based on publicly available information. It is not a republished article. Use the source link below for the original report.