Windsurf Pricing in 2026: What You Actually Pay Now

Windsurf pricing in 2026, in one table

If you searched for Windsurf pricing and landed on a different company's website, nothing is broken. The product still exists, it is still maintained, and you can still download it. It just has a different name and its pricing page moved.

Here is what you pay today, taken from Devin's live pricing page on September 18, 2026:

Those are the same four self-serve price points Windsurf published in March 2026. What changed is the brand on the invoice, the seat model underneath Teams, and where the page lives. The rest of this post covers how each of those happened, what the numbers actually buy, and what you should do if you are an existing subscriber.

Why windsurf.com/pricing sends you to devin.ai

This is the part most comparison posts still get wrong, so it is worth being precise about what the redirect actually does.

Requesting `https://windsurf.com/pricing` returns an HTTP 308 permanent redirect to `https://devin.ai/pricing`. The windsurf.com root does the same thing, landing on `https://devin.ai/desktop`. So does `/editor`, `/download`, `/blog` and `/enterprise`. A handful of deeper legacy paths still resolve on windsurf.com itself and serve the old Windsurf site chrome, which is why you occasionally still see the original branding, but every page a buyer would go looking for now terminates on Cognition's domain.

A 308 is the strongest signal a site can send that a resource has moved for good. It is not a temporary maintenance redirect, and it is not a marketing landing page swap. Cognition moved the canonical pricing for this product to devin.ai and left a permanent forwarding address.

The practical consequence: any Windsurf pricing figure you read that was not sourced from devin.ai or docs.devin.ai is, at best, a snapshot of a page that no longer exists. That includes most of the roundups currently ranking on this query.

The full timeline, with dates

The reason Windsurf pricing is confusing is that this product went through three separate ownership events in roughly seventy-two hours, and then a rename eleven months later. Here is the sequence with dated sources.

Three details in that sequence get misreported constantly, and each one changes how you should read any pricing claim you find elsewhere.

OpenAI never owned Windsurf. The $3 billion figure was widely reported in April and May 2025 and it was real, but the deal never closed. OpenAI's own spokesperson confirmed at the time that the exclusivity period had expired. If a post tells you Windsurf pricing is set by OpenAI, it is working from a headline about a deal that collapsed.

Google did not buy Windsurf either. Google hired the founders and paid for a non-exclusive licence. The company itself, including the product, the brand and the revenue, stayed independent for three more days. This structure is sometimes called a reverse acquihire, and the important consequence for a buyer is that Google has no say in what this editor costs.

Cognition bought the whole business, not just the code. Scott Wu's announcement lists what transferred: the IDE, the IP, the trademark and brand, $82M of ARR, 350+ enterprise customers and hundreds of thousands of daily active users, plus the team. That is why pricing decisions moved to Cognition immediately and why the product eventually took the Devin name.

What happened to the Windsurf brand

On June 2, 2026, Cognition renamed the Windsurf editor to Devin Desktop. Cognition's own FAQ is unusually direct about it: Devin Desktop is the new name for Windsurf, it is the same IDE with the same editor and the same features, unified under the Devin brand.

The stated reason is consolidation. Cognition now runs four surfaces under one name: Devin Desktop (the editor with an agent manager built in), Devin Cloud (the autonomous long-running agent on its own machine), Devin CLI (the terminal), and Devin Review (code review on every diff). Keeping a separate Windsurf brand for one of those four surfaces stopped making sense once the underlying agent was shared across all of them.

What actually changed in the product on that date, beyond the name:

• The Agent Command Center, which is a Kanban view of every local and cloud agent you are running, became the default surface instead of a secondary panel.
• Spaces launched, which share context and Git worktrees across multiple agents working on related tasks.
• Support for the Agent Client Protocol (ACP) landed, so third-party agents including Codex, Claude Agent and OpenCode run inside the same interface as Devin.
• Cascade, the local agent, was rewritten from scratch in Rust and renamed Devin Local. Cognition claims up to 30% better token efficiency and adds subagent support and sandboxing.

None of that changed the price. The rename FAQ states it three separate ways: your current plan continues to work exactly as it does today, pricing is unchanged, and that applies to all plan types including legacy Windsurf Enterprise plans.

What each plan actually buys

The headline numbers are easy. What separates the tiers is less obvious, and it matters more than the price difference for most buyers.

The Free plan, $0/month

The Free plan is a real product, not a countdown. It includes a light quota for agent work, unlimited inline edits and unlimited Tab completions, and access to what Cognition calls Fast Context.

What it does not include is more instructive. Per the plan comparison table on the pricing page, Free has no access to premium models, no Devin Cloud, no Devin API, no Ask Devin, no DeepWiki, and none of the Slack, Teams, Linear, Jira, GitHub, GitLab or Bitbucket integrations. Model availability is explicitly limited. Cognition's own SWE-2 model is available on Free but at reduced speed.

That combination makes Free a genuine evaluation tier and a workable one for a hobbyist who mostly wants autocomplete. It is not a plan a professional developer will stay on.

Pro at $20/month

Pro is the plan that maps most closely to what people mean when they say Windsurf Pro. It adds full model availability, described on the pricing page as access to frontier models from OpenAI, Anthropic and Google plus leading open source models. It adds Devin Cloud access, the Devin API, Ask Devin, DeepWiki, the full integration set, and the ability to purchase extra usage at API pricing once the included quota runs out.

It also currently includes free use of SWE-2, Cognition's own coding model, in Devin Desktop and the CLI through October 10, 2026. That is a dated promotional term sitting on the live pricing page, and it is worth noting precisely because it expires. If you are building a cost model past October 2026, do not assume SWE-2 stays outside the quota.

Pro is a single-user plan. Cognition's billing documentation is explicit that Pro subscribers cannot invite additional members to their organization.

Max at $200/month

Max is Pro with significantly higher quotas, and one structural difference that the marketing copy does not surface but the docs do: Pro has a daily and a weekly allowance, while Max has a weekly allowance only, with no daily cap.

That is a bigger deal than it sounds. A daily cap means a developer who wants to spend one long Saturday refactoring something large hits a wall regardless of how much weekly headroom they have left. Max removes that wall. If your usage is bursty rather than evenly spread, the absence of a daily cap can matter more than the raw quota increase.

Max is also single-user. It is not a way to cover two developers.

Teams at $80/month minimum plus $40 per full seat

This is the tier where the commonly quoted number is wrong, and it gets its own section below.

Enterprise

Enterprise is quoted rather than listed. Per the comparison table, it is the only tier that includes VPC deployment, SAML/OIDC SSO, centralized enterprise admin controls, teamspace isolation, custom git provider support, a dedicated account team and custom terms. Enterprise customers are billed in Agent Compute Units against a volume committed in their order form rather than on the self-serve quota system. Self-serve Teams caps at 200 members, so anything larger goes here by necessity.

The Teams plan is not simply $40 per user

Almost every Windsurf pricing page still in circulation says Teams costs $40 per user per month. That number came from Cognition's March 2026 announcement, where it was accurate. The seat model has since changed, and the current version is more interesting than the old one.

Per Cognition's self-serve billing documentation, every member of a Teams account holds one of two seat types:

A full seat costs $40/month, billed as a fixed recurring line item. It carries a daily and weekly usage quota equivalent to the Pro plan, and it is what grants access to Devin Desktop, which is to say the editor itself.

A flex seat is free. There is no separate cap on how many you have. Flex seats draw entirely from the team's shared pool of prepaid on-demand credits, carry no fixed monthly charge, and do not include Devin Desktop access.

On top of that sits an $80/month minimum for every Teams account. The docs spell out how it resolves:

So the $80 is a floor, not an additive platform fee. Below two full seats, the shortfall is charged as prepaid credits the whole team can draw on. At two full seats and above, you are simply paying $40 per full seat.

Note that the plan card on the pricing page phrases this as "$80/month for team plan + $40/mo per full dev seat", which reads like $80 plus $40 per person. The billing documentation and the comparison table both describe a minimum rather than a surcharge, and the docs include the arithmetic table above. If you are budgeting to the dollar, the docs are the version with the worked example in it.

Two things follow that the old "$40 per user" framing completely hides.

A small team pays more than the headline suggests. A two-person shop where only one person uses the editor pays $80, not $40. The extra $40 arrives as credits rather than vanishing, but it is still $80 leaving your account.

A mixed team pays far less than the headline suggests. The free flex seat is the single most useful fact in this pricing model and it is almost never mentioned. A ten-person engineering team where four people live in the editor and six dip in occasionally does not cost $400. It costs $160 in full seats plus whatever the six flex users draw from shared credits. The trade is that those six do not get Devin Desktop, so they are using Devin through the cloud, CLI or API surfaces rather than the IDE.

Get the seat mix wrong in either direction and you are off by a factor of two or more. This is the number to check before you renew.

How the quota actually works

Windsurf moved off credits on March 19, 2026. Cognition's stated reason was that flat-rate credits charged the same for a one-line question and a forty-minute agent session, which trained users to batch requests into oversized prompts and degraded the experience. The replacement is a usage allowance that refreshes on a calendar basis.

What consumes it, per the usage documentation:

• The number and complexity of actions the agent takes, which covers planning, context gathering, task execution, browser actions and code execution.
• Virtual machine time and network bandwidth, described as typically a small fraction of the total.

What does not consume it, which is the more useful half:

• Waiting for your response.
• Waiting for a test suite to run.
• Setting up and cloning repositories.
• Sleeping. Sessions sleep automatically after 30 minutes of inactivity by default and consume nothing while asleep. Enterprise admins can move that timeout between 5 and 120 minutes.

Two platform quirks are worth knowing before you budget. Windows sessions consume roughly 9% more than equivalent Linux sessions. macOS sessions currently consume the same as Linux, but Cognition flags that as promotional launch pricing subject to change.

Tab completion is unlimited on every plan including Free, and inline edits are unlimited on Free as well. The quota is aimed squarely at agent work, not at typing assistance.

What the quota is worth in messages

Cognition published estimated message volumes when it announced the change. These are estimates based on usage patterns rather than hard caps, and they were published in March 2026, so treat them as orders of magnitude rather than guarantees.

The spread inside each row is the honest part of this table. A Pro user can get anywhere from 8 to 101 Premium messages in a day depending on task size, context length and how much reasoning the model does. That is more than a 12x range on the same plan at the same price, which is why nobody can tell you what your bill will be without watching a month of your own usage.

The most reliable lever Cognition names for stretching a quota is model selection: routing routine work to small models and reserving frontier models for genuinely hard problems. The second is prompt discipline, being precise while cutting unnecessary context.

What happens when you run out

You buy on-demand credits, and usage past your quota draws from them at API pricing. Three properties of those credits are worth knowing:

• They roll over month to month and purchased credits never expire.
• They can be topped up manually or refilled automatically with an auto-reload threshold, and admins can set default per-session spending limits.
• On Teams they are shared across every member with no per-member balance, so any teammate can draw from the pool.

That last one cuts both ways. Shared credits mean an occasional user is never blocked by an artificial per-person limit, and they also mean one person can drain the pool for everyone. The per-session spending limit is the control that exists for this, and it is worth setting before you need it.

On Teams, Automations and Devin Review draw from the shared credit pool rather than from full-seat quota. If credits run dry, Automations stop running and Devin Review falls back to a plain diff viewer.

Worked cost scenarios

Every number below is arithmetic on the published prices. The assumptions are stated so you can swap in your own.

One developer who wants what Windsurf Pro was. Pro at $20/month. Full model access, daily and weekly quota, Devin Cloud, the full integration set. For a developer who runs agents regularly but not constantly, this is the plan, and there is nothing else to model.

One developer deciding between Pro and Max. Max costs $180/month more than Pro. So Max is the cheaper option only if you would otherwise buy more than $180/month in on-demand credits. Below that, Pro plus top-ups is less money. The exception is the daily cap: if your work is bursty and you regularly hit a daily wall on Pro while having weekly headroom left, Max buys you the ability to use your capacity when you want it, and that is worth something the arithmetic does not capture. Either way, this is a decision to make after a month of real usage data, not before.

A ten-person team where everyone lives in the editor. Ten full seats at $40 is $400/month, which clears the $80 minimum comfortably. That is the number the old "$40 per user" shorthand gives you, and in this specific configuration it is correct.

A ten-person team with four heavy users and six occasional ones. Four full seats at $40 is $160/month. The six occasional users take free flex seats and draw from shared credits. Assume those six consume $150/month of credits between them, and the total is $310/month against $400 for the all-full-seats configuration. The catch is real: those six do not get Devin Desktop, so if the editor is the point, they need full seats.

A solo developer who lands on Teams by mistake. One full seat at $40 plus $40 auto-charged as credits to clear the minimum is $80/month, for something Pro delivers at $20 plus $60 of credits. If you are one person, you want Pro, not Teams.

A forty-developer organization. Forty full seats at $40 is $1,600/month, or $19,200/year before any on-demand usage. Self-serve Teams caps at 200 members, so this is still inside the self-serve band, but at this size the Enterprise features that are gated behind a quote (SSO, VPC deployment, centralized admin controls, per-user usage policies) are usually the reason to talk to sales rather than the price.

Windsurf pricing against Cursor, Copilot, Claude Code and Codex

Every figure here was checked against the vendor's own live pricing page or documentation in September 2026.

Four observations that a straight price comparison misses.

At the entry tier, Windsurf and Cursor are identical on price and different on metering. Both are $20. Cursor's model gives you a dollar-denominated pool equal to your plan price that only drains when you manually pick a premium model, with its Auto mode running unlimited. Devin gives you a message-shaped daily and weekly allowance that every agent action draws against. Cursor's model rewards you for staying in Auto. Devin's rewards you for keeping sessions short and picking smaller models. Neither is better, but they punish different habits.

Copilot is the cheapest and the most forecastable. GitHub's plan documentation puts Pro at $10/month, Pro+ at $39, Max at $100, Business at $19 per seat and Enterprise at $39 per seat, with overage at a flat $0.01 per AI credit and code completions excluded from metering entirely. If your finance team wants one number to multiply by headcount, that is the model that gives them one. It buys you less control over which frontier model runs each request.

Anthropic publishes the only honest public benchmark for what agentic coding actually costs. Claude Code's cost documentation states that across enterprise deployments the average is roughly $13 per developer per active day and $150 to $250 per developer per month, with 90% of users staying under $30 per active day. Hold that against Devin Pro at $20 or Max at $200 and the picture gets clearer: a subscription that covers heavy agent usage inside a fixed monthly price is absorbing real cost, which is exactly why every one of these vendors meters the agent and gives away the autocomplete. Our fuller breakdown is in Claude Code pricing.

Codex does not have its own subscription. It is bundled into ChatGPT plans, so ChatGPT's pricing is Codex's pricing: a $0 Free tier, Go at $8, Plus at $20, Pro from $100, and Business at $20 per user per month billed annually or $25 monthly. OpenAI's Codex pricing documentation notes that ChatGPT Work and Codex share the same allowance, which means heavy chat usage eats into your coding budget. That is a structural difference worth catching before you standardize a team on it.

For a wider view of the category, see the best AI for coding and AI tools for software engineers.

What happened to existing Windsurf subscribers

Cognition handled this better than most acquisitions handle it, and the specifics are documented rather than implied.

Your price is grandfathered indefinitely. The March 2026 announcement states plainly that existing Pro and Teams users keep their current plan price indefinitely while migrating to the new usage allowance system. If you were paying a legacy Windsurf rate, you still are.

Free users were migrated automatically to the quota system with no action required.

Enterprise customers were unaffected in the immediate term and continue on existing billing agreements. The rename FAQ repeats this: pricing is unchanged for all plan types, including legacy Windsurf Enterprise plans.

Everyone got an extra week. Every existing paid subscriber received a free extra week on their current plan starting March 19, 2026, specifically so they could test the quota system before committing. Anyone unsatisfied after that week could cancel and request a prorated refund. That is an unusually confident way to ship a billing change, and it is worth noting that the company did it rather than simply announcing the switch.

Add-on credits were converted, not voided. If you had purchased add-on credits under the old system, they were converted into a dollar amount usable for extra usage, at a rate equivalent to what you paid for them.

The June 2026 rename changed nothing financial. It arrived as a standard over-the-air update. Plans, pricing, extensions, settings, workflows, keybindings, language servers and in-progress work all carried over. Windsurf rules files still work, and the new `.devin/` equivalents were added alongside rather than replacing them.

If you are on a legacy Windsurf plan, the docs note you can still use windsurf.com to manage your settings, which is part of why that domain has not been fully retired.

Is the editor still shipping and still maintained

Yes, with two qualifications that matter depending on how you use it.

The desktop editor is actively developed. Devin Desktop is the successor to Windsurf, not a maintenance branch. The June 2026 release added the Agent Command Center, Spaces, ACP support for third-party agents and the Rust rewrite of the local agent. Cognition describes it as where it expects developers to spend 90% of their day, which is not the language of a product being wound down. The editor, extensions, keybindings and LSPs remain backwards compatible with both Windsurf and VS Code.

Cascade has been replaced. The local agent is now Devin Local. Cascade remained available through July 2026 for incremental migration and is no longer the default. Pricing for Devin Local is identical to Cascade's, and it inherits your existing Windsurf settings.

The JetBrains plugin is the genuine casualty. Cognition's FAQ states that the Windsurf JetBrains plugin is in maintenance mode, that Cascade inside it is being deprecated, and that newer JetBrains IDE releases can break the legacy plugin. The download page still works and the plugin is still listed on the JetBrains Marketplace, but the recommended path is now running Devin over ACP inside JetBrains rather than using the native plugin. If your team standardized on IntelliJ or PyCharm, this is the part of the transition to plan for.

One small operational change: the `surf` shell command was replaced by `devin-desktop`. The legacy `surf` and `windsurf` binaries still ship in `~/.codeium/windsurf/bin/` for backward compatibility, and Cognition's own suggested fix is aliasing `surf` to the new command.

If you want off: what the alternatives cost

The rename is a reasonable trigger for re-evaluating, and there is no lock-in at the file level. The editor is VS Code compatible, so your extensions and settings port.

Staying put is the cheapest option if the editor works for you. Pro at $20 is competitive, your grandfathered price is protected, and the underlying product got more capable rather than less.

Cursor is the closest like-for-like swap. Same $20 entry price, same $200 top individual tier, a similar agentic-IDE model, and a credit system that rewards different habits. Team seats are $40 for Standard and $120 for Premium.

GitHub Copilot is the move if predictability is the priority. Pro at $10 halves your individual cost and Business at $19 per seat halves the team cost, with a flat $0.01 per credit overage and unmetered completions. You give up direct control over which frontier model handles each request.

Claude Code or Codex if you want the agent without the IDE. Both run in a terminal and inside other editors. Claude Code is $17/month on annual billing, Codex comes with a $20 ChatGPT Plus subscription you may already be paying for.

Nothing, if what you actually want is the free tier. VS Code with Copilot Free covers a surprising amount of ground for a hobbyist.

The honest note on switching: the cost of moving is not the subscription, it is the two weeks your team spends rebuilding muscle memory and rules files. Unless something is actually broken for you, a rename is a weak reason to pay that.

Who Windsurf pricing is now right and wrong for

Right for a solo developer who wants an agentic IDE at $20. Pro is priced identically to Cursor Pro and includes Devin Cloud access, which Cursor has no equivalent of. If you want one subscription covering both an editor agent and a cloud agent that works on its own machine, this is the only one of these tools that bundles both at $20.

Right for a mixed team with a small core of heavy users. The free flex seat is genuinely differentiated. No other vendor in this comparison lets you add occasional users at zero fixed cost. If four of your ten engineers live in an agent and six touch it weekly, this pricing model is built for you and the competition's per-seat pricing is not.

Right for teams that want one agent across editor, terminal, cloud and code review. Full-seat quota works across Devin Cloud, Devin Desktop, Devin CLI and Devin Review. Buying four separate tools to cover those surfaces costs more.

Wrong if you need a fixed, forecastable line item. Quota plus on-demand credits is the correct model for the underlying economics and a hard model to forecast in your first quarter. Copilot Business at $19 with a fixed credit allowance is the more honest fit for a budget owner who needs a number that does not move.

Wrong if your team is standardized on JetBrains. The native plugin is in maintenance mode with its agent deprecated. ACP works, but you are on a migration path rather than a supported steady state.

Wrong if brand stability matters to your procurement process. This product has been Codeium, then Windsurf, then Windsurf-under-Cognition, then Devin Desktop, inside two years. Nothing about that has hurt users so far, and Cognition has handled each transition with grandfathered pricing and automatic migrations. But if you are writing a three-year vendor plan, it is a fact to weigh rather than ignore.

Wrong if you want the cheapest autocomplete. Copilot Pro at $10 covers unlimited completions for half of Devin Pro's price. If agent runs are occasional and Tab is the real value, you are overpaying here.

Where a coding assistant sits in a wider AI budget

A coding tool is one line in a larger AI spend that usually also covers a general-purpose assistant, whatever models a team calls directly, and increasingly a context layer connecting those tools to the rest of the company's systems. Our enterprise AI pricing comparison and the enterprise AI price index cover how those categories stack up, and enterprise AI engineering workflows covers where the spend actually earns its keep.

Where Coworker fits alongside a coding tool specifically: Devin Desktop makes writing code faster inside the editor. Coworker MCP gives whichever coding tool you use unified context across the systems a real engineering task touches, with 50+ connectors spanning the Jira ticket describing the bug, the Slack thread where a customer reported it, and the history of the file being changed. These are adjacent problems, not competing ones, which is why Coworker works the same way whether your team standardizes on Devin Desktop, Cursor, Copilot or Claude Code.

If you are evaluating that layer, how to choose MCP servers for your team covers the criteria, what is MCP covers the protocol itself, MCP security covers the governance questions that come up when AI tooling touches production systems, and enterprise MCP deployment covers rolling it out. For the coding-specific version, Claude Code MCP servers is the closest analogue.

Book a demo to see how Coworker MCP connects to the coding tools your team already uses.

Frequently asked questions

Is Windsurf still a separate product?

No. Windsurf was renamed Devin Desktop on June 2, 2026. It is the same IDE with the same editor, extensions and features, now under the Devin brand alongside Devin Cloud, Devin CLI and Devin Review. The product still ships and is still actively developed.

Where did the Windsurf pricing page go?

As of September 18, 2026, windsurf.com/pricing returns a 308 permanent redirect to devin.ai/pricing. The windsurf.com root redirects to devin.ai/desktop. A 308 is permanent, so devin.ai is now the canonical source for these prices.

How much does Windsurf cost in 2026?

Free is $0, Pro is $20/month, Max is $200/month, and Teams has an $80/month minimum with full seats at $40/month each. Enterprise is priced on request. Those are Devin's published plans, and they cover Devin Desktop, which is what Windsurf is now called.

Did my Windsurf price change after the acquisition or the rename?

No. Existing Pro and Teams subscribers were grandfathered at their current price indefinitely when the quota system launched in March 2026, and the June 2026 rename explicitly changed no pricing for any plan type, including legacy Windsurf Enterprise plans.

Is Windsurf Teams still $40 per user per month?

Not exactly, and this is the figure most posts get wrong. Full seats are $40/month and do include Devin Desktop access, but flex seats are free and draw from shared credits instead, and every Teams account pays at least $80/month. A team with one full seat pays $80, and a team with four full seats and six flex seats pays $160 plus shared credit usage rather than $400.

Is there still a free version of Windsurf?

Yes. The Free plan costs nothing and includes a light agent quota, unlimited Tab completions and unlimited inline edits. It has limited model availability and no access to premium models, Devin Cloud, the API or the Slack, Jira, Linear and GitHub integrations.

What happened to my Windsurf credits?

Credits were retired on March 19, 2026 in favour of daily and weekly quotas. Add-on credits purchased under the old system were converted into a dollar amount usable for extra usage, at a rate equivalent to what you paid for them.

What happens when I hit my usage quota?

You buy prepaid on-demand credits, and usage past your quota draws from them at API pricing. Credits roll over month to month and never expire. On Teams they are shared across the whole team with no per-member balance, and admins can set auto-reload thresholds and per-session spending limits.

Is Max worth $200 over Pro at $20?

Only if a month of real usage says so. Max costs $180 more, so it is the cheaper option only if you would otherwise buy more than $180/month of on-demand credits. The structural argument for Max is separate from the arithmetic: Pro has a daily and a weekly cap while Max has a weekly cap only, so Max suits bursty work where you want to spend a full day on one large task.

Does the Windsurf JetBrains plugin still work?

It is still downloadable, but Cognition has put it in maintenance mode and is deprecating Cascade inside it, and warns that newer JetBrains IDE releases can break the legacy plugin. The recommended path is running Devin inside JetBrains over the Agent Client Protocol instead.

Who owns Windsurf now?

Cognition, the company behind the Devin coding agent. It acquired Windsurf's IP, product, trademark, brand, business and team on July 14, 2025, three days after OpenAI's $3 billion acquisition collapsed and Google hired the founders in a $2.4 billion licensing deal. OpenAI never owned Windsurf and neither did Google.

Do I need Devin Cloud to use the editor?

No. You can run Devin Desktop with local-only agents. Devin Cloud is included from Pro upward on self-serve plans, and remains a separate SKU for Enterprise customers.

What counts against the quota, and what does not?

Agent actions count: planning, context gathering, task execution, browser actions and code execution, plus a small amount for VM time and bandwidth. Waiting for your response, waiting for a test suite, cloning repositories and sleeping sessions do not. Tab completions and inline edits are unlimited on every plan and never draw from the quota.

Related reading

• Claude Code Pricing
• The Best AI for Coding
• Enterprise AI Pricing Compared
• How to Choose MCP Servers for Your Team
• What Is MCP? A Complete Guide