Windsurf Is Now Devin Desktop: What Actually Changed
Windsurf is the former name of Devin Desktop, the editor Cognition unified with Devin Cloud, Devin CLI and Devin Review on June 2, 2026. Cognition acquired Windsurf's product and team in July 2025, never disclosed the price, and replaced the Cascade agent with Devin Local, a Rust rewrite it says is up to 30% more token efficient.
The short version
Cognition renamed Windsurf to Devin Desktop on June 2, 2026, by over-the-air update. Plans and prices carried over, but the default screen became an Agent Command Center and Cascade, the original agent, is now labelled legacy in the docs. Re-checked on 30 October 2026, no removal date has been published.
Not sure which one fits you? Describe your situation and Smart Match names three tools.Find my AI tools
In this guide9 sections
On June 2, 2026, every existing Windsurf install turned into something called Devin Desktop, and the maker did not ask first.
The change shipped as a regular over-the-air update, the way a phone gets a patch. No uninstall. No migration wizard. Cognition's FAQ says users get a heads-up that the experience is changing, but nothing in it offers an opt-out. A developer who had Windsurf open that morning had a renamed app by lunch, with the same login and the same files.
That matters to anyone who evaluated the editor in the past year, because many of the pages that still rank for its name describe an independent product, usually written before the July 2025 acquisition by Cognition and long before the rename that changed what the app shows on launch and what its maker calls it.
It has not been independent since June. Not in name, anyway. Almost five months on, this guide re-checks what changed, what the vendor has since put in writing, and what a buyer should do about it. Where the vendor makes a claim we cannot test, the text says so.
Hokai timeline of Windsurf's rename, from TechCrunch's deal report, Cognition's FAQ and Devin's plan docs as re-read on 30 October 2026.
What Changed on June 2
The headline change is the default view. Windsurf used to open on the code editor, with an assistant available on request. The renamed app opens on the Agent Command Center, a Kanban board where every local or cloud session gets a card sorted by status. The editor still exists. It is now a tab inside a fleet manager, rather than the other way around.
Underneath, the company replaced Cascade, the agent that built the product's reputation, with Devin Local. It says the new agent was rewritten from scratch in Rust, is up to 30% more token efficient, and supports subagents. Those are vendor claims. The announcement publishes no benchmark behind the 30% figure, so treat it as a design goal until someone measures it.
The app also gained support for the Agent Client Protocol, an open standard that lets outside agents run inside the same editor. At launch the vendor named Codex, Claude Agent and OpenCode. Third-party agents get the same board and the same shared context as Devin's own, which is the real strategic move: the product is now a place to manage agents, whoever built them.
| Area | Before June 2 | After |
|---|---|---|
| Default screen | Code editor | Agent Command Center |
| Local agent | Cascade | Devin Local |
| Outside agents | Not supported | Via ACP |
| Plans and prices | Free, Pro, Max, Teams | Unchanged |
| Terminal command | surf | devin-desktop |
The old surf command still ships for backward compatibility. Existing .windsurfrules files are still read, and the new .devin/rules/ folder takes precedence when both exist, according to the FAQ. The same document puts the JetBrains plugin in maintenance mode and tells those users to connect Devin over the protocol instead.
Inside the Agent Command Center
Short version: it is a task board for software agents. Each session, whether it runs on your laptop or in the cloud, becomes a card. The Kanban layout lets a developer start four tasks and come back to whichever one needs a human.
A second feature, called Spaces, groups related sessions, pull requests, files and context in one place so that different agents can share what they know. The announcement describes it as a way to share context between agents. It does not describe limits on how much context a Space can hold, which matters for large repositories.
The vendor says the editor is still a full IDE underneath, backward-compatible with Windsurf and VS Code extensions, language servers and workflows. If that holds, switching costs for a current user are close to zero. The cost shows up elsewhere: attention. A board with many cards invites more parallel work, and parallel work needs more review.
For a code reviewer, that is the real trade. More agents mean more diffs. Teams that adopt the board without changing how they review will find the queue, not the typing, is what slows them down.
Who It Affects
Existing users got the better end of this deal, at least on paper. The update cost nothing and broke nothing the vendor has disclosed. Its FAQ says plans and pricing carry over, and the pricing page as of 30 October 2026 still lists the same ladder.
Devin's self-serve plans documentation, captured 19 Aug 2026. The pricing page was re-read as text on 30 Oct 2026 and lists the same ladder.
| Plan | Price | What the vendor lists |
|---|---|---|
| Free | $0 | Light quota, unlimited Tab |
| Pro | $20 a month | Higher quotas, free SWE-2 |
| Max | $200 a month | Much higher quotas |
| Teams | $80 plus $40 per seat | Up to 200 users |
| Enterprise | Custom | SSO, dedicated deployment |
Cognition benefits too, and more directly. Folding an editor with an established user base into the same brand as Devin turns each session into a funnel toward the cloud agent. At launch the FAQ said Devin Cloud remained a separate product. The pricing page now lists cloud-agent access under Pro, which makes that funnel shorter than it was in June. Cursor makes a similar pitch with its own background agents, minus the rebrand.
Teams buyers should read one line closely. Devin Desktop is available to full seats only. A full seat carries $40 a month of usage that works across all the vendor's products. So a ten-person team pays the flat plan fee once and a seat fee ten times. That is simple, and still not the number a casual reader of an old pricing post would expect.
How Usage Works Now
Every paid plan comes with an allowance that refreshes on a daily and a weekly cycle. The vendor says most users never hit it. Those who do can buy extra usage, billed at the underlying model's API price. Cost per message varies with the model, the size of the task and the reasoning effort, so two developers on the same plan can burn quota at very different speeds.
The Free plan is the odd one out. It offers a light quota for agents and limited model availability, but inline edits and Tab completions are unlimited. That makes it a fair way to test the editor and a poor way to judge the agents.
Switching to the new local agent does not change how you pay. The FAQ says Devin Local uses the same pricing model, so anyone on prompt-based credits stays on them. The cheap way to cut spend is also the dull one: shorter prompts, less pasted context, and smaller models for routine edits. The vendor's own pricing FAQ recommends the same three habits.
Who Gets Left Behind
Four groups lose out, and none of them shows up in the announcement.
- Readers of stale search results. Their comparison pages describe an editor that no longer exists in that form.
- People who chose an editor first. The classic view is still reachable, but it is no longer the default.
- Cascade loyalists. Legacy access was promised through July 1. New chats now start in Devin Local.
- JetBrains users. The plugin is in maintenance mode, and newer JetBrains releases can break it.
The first group is the largest and the hardest to reach. A team that picks tools from a year-old post is comparing against a product that has been folded into something else. That is the reason this page is dated and re-checked rather than left alone.
Cognition explains the rebrand as unifying Devin Cloud, Devin Desktop, Devin CLI and Devin Review under one name. That is a branding choice. Rivals such as GitHub Copilot, from GitHub, did not need to make it to stay competitive, and the cost of the choice lands on people who never asked for it, from solo developers who now search for a name that no longer ships to IT teams who must review a new set of hostnames and file paths.
The Case for Staying Put
The obvious objection: pricing did not change, compatibility did not break, and the editor sits on the same foundation it always did. By that measure this is a UI update wearing a rebrand's clothes, and calling it news overstates it.
There is a fair version of that argument. A developer who wanted a fast editor with a competent agent still has one. Tab completions stay unlimited on every plan, existing rules still load, and the vendor says nothing was removed. Switching tools has its own cost in habits, shortcuts and review time, and churn for a name change would be a poor trade.
The argument stops working at the moment someone has to buy or recommend the tool. Asking a colleague whether to use the old name gets an answer about a product the vendor has retired from everything new it builds.
And the agent that earned the original its reputation is now labelled legacy in the vendor's own documentation. A UI update does not usually relabel the part of the product people came for. So the sensible position is to stay, with open eyes and a dated plan for the agent change.
The Money and the Data Angle
Cognition has never disclosed what it paid. TechCrunch's report on the July 2025 deal says plainly that the price was not announced. It does record that the editor had reached $82 million in annual recurring revenue and at least 350 enterprise customers by Cognition's own count. It also notes that Google had hired the CEO and a co-founder days earlier in a $2.4 billion reverse-acquihire.
For a buyer the practical questions are narrower. Data handling is one. The FAQ documents the hostnames the app contacts for updates and remote development, the local paths it reads and writes, and the policies that endpoint-security teams should allow. An IT lead who has to approve the new name should read those lists before the next renewal, because the old allowlist will not cover them.
Contrast this with Qodo, an AI code-review platform. In April 2026 it moved its open-source PR-Agent into a community-owned GitHub organization and restored the Apache 2.0 license, according to analyst coverage dated 23 April. Two vendors facing the same pressure to make AI-written code trustworthy chose opposite directions: one folded a product deeper into its brand, and the other gave a tool away.
The pattern is worth naming. A renamed product leaves a gap between what buyers remember and what ships. Nobody lied here, and nothing broke. But a decision made from memory of the old name is a decision about a product that has moved, which is why a dated check beats a remembered one.
What to Do About It
- Check which plan you are on. A legacy plan keeps working unchanged, per the FAQ.
- Open the app once and confirm that rules, extensions and keybindings carried over.
- Decide whether the Agent Command Center should be your home screen, and pick agents on purpose rather than by default.
- If you still use Cascade, plan a move to Devin Local while Cascade remains available.
- Compare alternatives using current prices, not last year's posts.
For that comparison start with our best AI coding assistants roundup and the AI code review tools guide, both re-checked on a schedule. If your stack is unusual, Smart Match narrows the field from a few questions about what you build.
On a tight budget, begin with the free coding options and open-source tools such as Aider. Other assistants worth a look are Trae, from ByteDance, and Tabnine, and the full list lives in the tools directory.
Builders who want apps rather than code can look at Replit or Lovable, which solve a different problem. To place Cursor, read what Cursor is for and the head-to-head Cursor Composer 2 against Claude Code. A related read on the model side is Grok Build and Cursor's Composer models.
What Happens to Cascade Next
Re-checked 30 October 2026: the announcement's July 1 date has passed, and the documentation still describes Cascade as available but legacy. The first version of this guide set a test for September 30, 2026, asking whether the company would say anything about Cascade's fate. No removal date appears in the documentation today.
What can be said is limited. Cascade is no longer the default. New work starts in Devin Local. Existing Cascade conversations stay available from history, and the legacy page still explains how to open it. The next signal to watch is the documentation itself. When the Cascade page disappears or gains a retirement date, that will be the announcement. Until then, treat Cascade as a feature with no promised future and plan around it.
For the wider market, see our agentic AI guide, the roundup of local coding models, and the earlier piece on how Devin was split into three products. The Windsurf and Devin records carry the plan details we re-verify.
Frequently asked questions
Is Windsurf still called Windsurf?
No. Cognition renamed it Devin Desktop on June 2, 2026, and delivered the change to existing installs as an over-the-air update. Some of its documentation still uses the Windsurf name for the classic editor view and for legacy plans managed at windsurf.com.
Did Windsurf's pricing change when it became Devin Desktop?
No. Cognition's FAQ says plans and pricing are unchanged, and its pricing page as of October 30, 2026 lists Free at $0, Pro at $20 a month, Max at $200 a month, and Teams at $80 a month plus $40 per full seat. Existing extensions, rules and keybindings carried over without a migration step.
What happened to Cascade, Windsurf's original AI agent?
Cognition replaced it with Devin Local, a Rust rewrite the company says is up to 30% more token efficient. The announcement promised legacy Cascade access through July 1. Cognition's documentation now labels Cascade the legacy agent, says new conversations start with Devin Local, and gives no removal date.
Can I still use Windsurf like a regular code editor?
Yes. Cognition says the classic editor, extensions, keybindings and workflows remain fully accessible inside Devin Desktop. The difference is that the app now opens on the Agent Command Center by default instead of the code editor.
How much did Cognition pay to acquire Windsurf?
Cognition has never disclosed the purchase price for its July 2025 deal for Windsurf's product and team. TechCrunch's report on the deal says the price was not announced, and records that Google separately hired Windsurf's CEO in a $2.4 billion reverse-acquihire days earlier.
Covered in this guide
- Devin: Devin is Cognition's autonomous AI software engineer that plans, codes, tests, and ships pull requests, with a Free plan and several paid tiers.
- Windsurf: Windsurf, now Devin Desktop from Cognition, is a VS Code-based editor with local and cloud agents, free with unlimited Tab and Pro at $20 a month.
- Aider: Free, open-source terminal pair programmer that edits your git repo and commits each change; bring your own LLM key. 49,000+ GitHub stars.
- Anthropic: Anthropic is a Public Benefit Corporation started by 7 ex-OpenAI researchers. It builds the Claude model family and counts Amazon and Google among its largest backers.
- ByteDance: ByteDance is a Beijing-based consumer-internet group whose short-video apps and Seed AI models power the Doubao assistant.
- Cursor: Cursor is an AI code editor built on VS Code, widely adopted across large enterprise engineering teams, with Agent Mode, Tab completion, and Cloud Agents.
- GitHub: Founded in 2008, GitHub is the largest code repository and collaboration platform with 140M+ developers, now integrated into Microsoft's CoreAI division after CEO Thomas Dohmke departed end of 2025.
- GitHub Copilot: GitHub's AI coding assistant with agent mode, Copilot Fusion model routing and MCP support across 6 IDEs, plus a free plan with no credit card required.
- Google: Google (Alphabet, NASDAQ: GOOGL) runs Search, Cloud and the Gemini models, including Gemini 4 Argon (October 2026).
- Lovable: Build something lovable. Create apps and websites by chatting with AI.
- OpenAI: OpenAI builds ChatGPT, the GPT-6 family (Astra, Sol, Terra, Luna), Codex and the OpenAI API. It reports 1.2B weekly ChatGPT users and raised new funding in March 2026.
- Replit: Cloud coding environment where you describe an app in plain English and AI builds and deploys it.
- Tabnine: Enterprise AI coding assistant with privacy-first architecture, agentic workflows, and flexible deployment
- Trae: Free AI IDE by ByteDance with SOLO autonomous agent and 4 plans from $3/month; supports Claude 3.7 Sonnet, GPT-4.1, Gemini 2.5 Pro, and DeepSeek on macOS, Windows, Web, and iOS.
Sources
Still deciding?
This guide covers a handful of options. Smart Match checks every listing in the directory against how you actually work and what you can spend, then hands you the shortlist and the reason behind each pick.
Find your tools

