Is Devin Desktop Free?
Free for the local editor, paid for the cloud agents. Cognition publishes a free plan that covers the Devin Desktop download with unlimited Tab completions, unlimited inline edits, a light quota to code with agents and limited model availability, while the Devin Cloud layer behind the editor starts on Pro.
One note on sourcing belongs at the top rather than buried at the bottom. The plan names and prices below were read off the Cognition pricing page on 17 September 2026. A few numbers people reasonably want are still absent from it, and those are listed further down rather than guessed at here.
That matters more than usual here, because this product replaced another one mid-year and the pricing surface moved with it. AI Perks tracks the credit programs that offset tools in this category.
Devin Desktop Pricing as Published
Cognition publishes a free plan plus four paid tiers: Pro at $20 per month, Max at $200 per month, Teams at $80 per month for the team plan plus $40 per month per full dev seat, and an Enterprise plan quoted on request rather than at a list price.
| Plan | Published price | What Cognition lists on it | Source |
|---|---|---|---|
| Free | $0 | Local Devin Desktop download, unlimited Tab completions, unlimited inline edits, light quota to code with agents, limited model availability | Pricing page, checked 17 Sep 2026 |
| Pro | $20 / month | Everything in Free, plus increased quotas, full model availability, cloud agents (Devin Cloud), DeepWiki, Ask Devin, the Devin API, extra usage at API pricing | Pricing page, checked 17 Sep 2026 |
| Max | $200 / month | Everything in Pro, plus significantly higher quotas and unlimited concurrent sessions where Pro caps at 10 | Pricing page, checked 17 Sep 2026 |
| Teams | $80 / month plus $40 per full dev seat | Everything in Pro, plus unlimited team members, sharing and collaboration, centralised billing, admin dashboard with analytics, priority support | Pricing page, checked 17 Sep 2026 |
| Enterprise | Custom, quoted on request | Dedicated deployment option, SAML/OIDC SSO, centralised admin controls, dedicated account and engineering support. Earlier reporting put compute unit billing near $2.25 per ACU, a figure the current pricing page does not carry | Pricing page plus older reporting, low on the ACU number |
Two things in that table are worth reading carefully.
Unlimited Tab sits on the free row. Autocomplete and inline editing are not the upsell in this product. They are the acquisition channel, and the metered thing is the cloud agent work.
The gap between Pro and Max is a 10x price step for quota, not for capability. Cognition describes Max as everything in Pro plus significantly higher quotas, with unlimited concurrent sessions where Pro caps at 10. Nothing in the feature list is exclusive to Max, so the only reason to pay the difference is a Pro quota or a concurrency cap you are genuinely hitting. See what the paid tiers look like after credits at AI Perks.
What Devin Desktop Actually Is
Devin Desktop is a unified software development environment that combines a code editor with an agent management layer for coordinating multiple Devin agents across projects. Cognition states plainly that Devin Desktop is the new name for Windsurf, and windsurf.com now returns an HTTP 308 redirect to devin.ai/desktop.
The agent management layer is the part that distinguishes it from an ordinary editor. Rather than one assistant inside one file, the framing is a console for running several agents across several projects at once, with the editor as the place you review what comes back.
The redirect is the clearest signal about Windsurf. When a product domain points at the successor product's own page, the old brand has stopped being a separate destination. Cognition answers the obvious follow-ups on that page: the change arrives as a standard over-the-air update, plans, pricing, extensions and settings carry over automatically, legacy Windsurf Enterprise plans keep their existing terms, and Windsurf for JetBrains (IntelliJ IDEA, PyCharm, WebStorm and others) is still available to download. In Cognition's framing, only the name and the branding changed.
There is also an earlier and more weakly sourced data point. Around April 2026, an integration was reported that let a developer plan work locally in Windsurf, hand it to Devin in the cloud, and review the returned pull request back in the IDE. Pricing, free tier terms and credit allowances for that integration were never published. That report carries low confidence, and Desktop now unifies both halves anyway, so read it as historical context rather than a live product line.
The Part That Costs Money Is the Cloud Layer
The split that matters is local versus cloud. Tab completions and inline edits run on your machine and are unlimited on the free plan. Cloud agents, DeepWiki, Ask Devin and the Devin API run on Cognition's infrastructure and start on Pro.
That is a coherent business model rather than an arbitrary line, and it tells you how to plan usage. The editing loop you run hundreds of times a day carries no marginal cost under the free plan. The delegated work that consumes someone else's compute is where the meter sits.
At the Enterprise end, earlier reporting described billing in agent compute units rather than a per seat subscription, with a unit put at roughly $2.25 and characterised as about 15 minutes of work. The current pricing page carries no ACU figure at all, and describes paid plans in terms of daily and weekly usage allowances instead, so treat the ACU number as stale third-party sourcing rather than a quotable price and ask for a written quote.
| Layer | Where it runs | Cost posture |
|---|---|---|
| Tab completions | Local | Unlimited on the free plan |
| Inline edits | Local | Unlimited on the free plan |
| Cloud agents | Cognition infrastructure | Paid, starts at Pro |
| DeepWiki, Ask Devin, Devin API | Cognition infrastructure | Paid, starts at Pro |
| Enterprise workloads | Cognition infrastructure | Custom, quoted on request, no published ACU rate |
AI Perks keeps credit programs for the model and infrastructure layers that sit underneath workflows like this one.
What Has Not Been Published
Being explicit about gaps is what keeps a pricing page from quietly rotting. Cognition now publishes more than launch coverage did, and a few numbers people reasonably want are still absent.
Free plan quotas in units. Cognition says the free plan carries a light quota to code with agents and limited model availability, on top of unlimited Tab completions and unlimited inline edits. The light quota is never expressed as a number of runs or messages, and cost per message is described as varying with the model, the task size and the reasoning required, so the free ceiling cannot be modelled in advance.
Credit or trial allowances. There is one published promotion rather than none: SWE-2, the in-house model, is listed as free in Devin Desktop and the CLI through 10 October 2026. Past that date no replacement rate is published, and no credit grant or trial length is published for anything else.
How large the Max quota actually is. What Max adds is now published, and it is quota rather than features: everything in Pro, plus significantly higher quotas and unlimited concurrent sessions where Pro caps at 10. Neither quota is given a number, so a 10x price step cannot be checked against a 10x anything.
Annual, student or nonprofit rates. Only monthly figures appear on the pricing page. The absence of an annual rate there is not proof that none exists on a sales call.
Platform coverage for the desktop build. Which operating systems the Desktop download supports is not stated on the pricing page. Windsurf for JetBrains is confirmed as still downloadable, which is a different question from which desktop operating systems the main build ships on.
What overage actually costs. The mechanism is published: go past the included usage on a paid plan and you can purchase extra usage, consumed at API pricing. Which API price list applies, per model, is not published, and no overage path is described for the free plan at all.
How to Keep the Cost Down
Start with credits on the layers underneath the editor. Coding agents and the model APIs they call are two bills that move together, and the second one is usually larger. Filter the AI Tool and Developer Tools categories at getaiperks.com for programs that cover that side of the stack.
Then exhaust the free tier deliberately. Unlimited Tab and inline edits mean a large share of daily editing carries no marginal cost. Reserve delegated agent work for tasks you would otherwise hand to a person, which is where the economics actually favour paying. Note the date on the free SWE-2 window too: it runs in Devin Desktop and the CLI through 10 October 2026, so anything you planned to try on the in-house model is cheaper this month than next.
Price Pro before Max. The higher tier buys quota, not capability, so the only thing that justifies a 10x step is a Pro quota you are actually exhausting or a concurrency cap you are actually hitting.
Finally, re-check before renewal. This product replaced Windsurf inside a single year and its pricing surface moved with it. Anything you model today deserves a fresh look next quarter.
Frequently Asked Questions
Is Devin Desktop free to download?
Yes. Cognition publishes a free plan covering the local Devin Desktop app, with unlimited Tab completions, unlimited inline edits, a light quota to code with agents and limited model availability. Cloud agent features start on Pro at $20 per month. The free plan carries no published expiry, and the light quota is not expressed as a number of runs or messages. Credit programs for the wider AI stack are tracked at getaiperks.com.
What happened to Windsurf?
Cognition renamed it. In its own words, Devin Desktop is the new name for Windsurf, built on the Windsurf IDE foundation with an agent command centre added on top. The windsurf.com domain now returns an HTTP 308 redirect to devin.ai/desktop. Existing users receive it as a standard over-the-air update, with plans, pricing, extensions and settings carried over, legacy Windsurf Enterprise plans keeping their terms, and Windsurf for JetBrains still available to download.
How much does Devin Desktop cost per month?
$0 for the free plan, $20 per month for Pro, $200 per month for Max, and $80 per month for the team plan plus $40 per month per full dev seat. Enterprise is quoted on request with no list price. Those figures were read off the Cognition pricing page on 17 September 2026. See what credits can offset at getaiperks.com.
What is the difference between Devin Desktop and Devin Cloud?
Devin Desktop is the local application: a code editor plus a layer for managing multiple agents across projects. Devin Cloud is the hosted agent service, covering cloud agents, DeepWiki, Ask Devin and the Devin API. The split is that Desktop has a free plan and the Cloud features begin on Pro.
Does Devin Desktop charge per ACU?
Not on the published plans. The pricing page describes daily and weekly usage allowances on each paid plan, with extra usage purchasable at API pricing, and it does not mention ACUs anywhere. Earlier reporting put an agent compute unit near $2.25 and about 15 minutes of work, a figure no current Cognition page carries, so ask for a written quote rather than relying on it.
Can free AI credits reduce what I pay for coding agents?
Credits do not pay a Devin subscription directly, and no credit program for Devin or the former Windsurf is published. The one vendor giveaway on the page is free SWE-2 use in Devin Desktop and the CLI through 10 October 2026. What credits do cover is the model APIs, compute and infrastructure surrounding an agent workflow, which is frequently the larger bill. An agent with no funded API budget behind it stalls quickly. AI Perks tracks credit programs across those layers at getaiperks.com.
Let the agents write the code. Let someone else fund the stack.