Windsurf Pricing (2026): Buy Pro, Meter the Rest

Windsurf Pro is the only defensible default at $20/month: Free is viable for autocomplete and light agent work, while Max jumps to $200 without publishing the quota that would justify the extra $180. For a five-developer team, five full seats cost $200/month, twice five separate Pro subscriptions, so Teams is a governance purchase rather than a usage discount.

Windsurf pricing at a glance

Windsurf costs $0, $20, or $200 per month for an individual. Teams starts with an $80 monthly minimum and converts to $40 per full developer seat once two seats satisfy that minimum. Enterprise pricing is custom.

The live Windsurf pricing URL now serves Cognition's unified Devin Plans and Pricing page. The subscription covers Windsurf, now described in the billing documentation as Devin Desktop, alongside Devin CLI and cloud-agent usage. Prices and billing mechanics below were verified against Windsurf's live pricing page and Devin's self-serve billing documentation on 13 August 2026.

The table contains every public list price, but it cannot contain the most important allowance: Windsurf does not publish the exact Free, Pro, Max, or Teams quotas. Its pages describe them as light, increased, or significantly larger. That omission prevents a buyer from calculating how many agent sessions the subscription includes before paying.

Who can use Windsurf Free without paying

Windsurf Free is a permanent $0 editor for autocomplete, inline changes, and light agent work. It is not a dependable agent budget for someone who codes with frontier models throughout the workday.

Free includes unlimited Tab completions and unlimited inline edits. Tab predicts code as you type; inline edits change a selected piece of code from a direct instruction. Those two workflows do not need a paid plan, which makes Free durable for developers who mainly want a smarter editor rather than an autonomous coding agent.

Agent use sits behind a light daily and weekly quota. The exact allowance is not published. Model access is also limited, and the current model documentation says Free users have access only to SWE-1.6. When the included quota runs out, a Free user cannot buy an overage and continue immediately. The documented action is to wait for the next daily or weekly reset.

That makes the honest free-plan boundary simple:

• Stay free indefinitely if Tab and inline edits do most of the work, agent prompts are occasional, and waiting for a reset has no business cost.
• Use Free as an evaluation plan if you need to inspect Windsurf's codebase context and agent workflow before adopting it.
• Do not rely on Free for delivery work if a blocked agent session can delay a release, client handoff, incident fix, or paid build day.

A solo builder maintaining a side project may never need to pay. The editor can keep completing and changing code even when the agent allowance is exhausted. A founder using an agent for multi-file implementation every day should budget $20 from the start. The difference is not the sophistication of the project. It is whether waiting until a reset is acceptable.

Windsurf sometimes offers trials of paid plans, but its documentation is unusually direct about the limitation: trials are promotional, appear only for a subset of eligible customers, and cannot be added retroactively by support. Free is the reliable $0 offer. A trial is not part of the plan.

Pro vs Max: the $180 decision

Buy Pro first. Max removes Pro's daily cap and raises the weekly quota, but Windsurf withholds the actual allowance on both plans. Without those values, the only defensible Max decision comes from your usage meter and the cost of interruption.

What $20 Pro buys

Pro is the working plan for one developer. It adds daily and weekly agent quota, full model availability, frontier models from OpenAI, Anthropic, and Google, cloud agents through Devin Cloud, and paid usage after the allowance is gone. The plan also includes Slack, Linear, and MCP integrations. MCP, or Model Context Protocol, is the connection layer that lets an agent use approved external tools and data.

Windsurf says the daily quota is more than one-seventh of the weekly quota. That matters because a developer who works heavily on two weekend days can still consume a useful share of the weekly allowance. The daily limit remains a hard pacing mechanism, however. A large migration on Tuesday cannot borrow all of Friday's unused allowance.

Pro also lets you keep working with on-demand credits after the included quota runs out. Extra usage is billed at the selected model's API list price. Purchased credits roll over from month to month and never expire. The subscription therefore buys an undisclosed allowance plus access to a metered continuation path.

What $200 Max changes

Max is still a one-person plan. It includes everything in Pro, replaces the daily-and-weekly shape with a significantly larger weekly quota, and removes the daily cap. That change suits bursty work: a long refactor, migration, or agent-heavy build can consume usage on the day it is needed rather than waiting for tomorrow's reset.

The missing number is decisive. Windsurf does not state how much larger the weekly quota is, in tokens, dollars, or a stable number of requests. “Significantly higher” cannot be compared with the $180 monthly premium. It also cannot be normalized against Cursor Ultra's disclosed usage pool or GitHub Copilot Max's disclosed AI credits.

Max can still be rational. It is rational when Pro repeatedly interrupts valuable work and when purchasing API-priced overage creates more operational friction than the upgrade removes. It is not rational merely because an agent is used every day.

What the overage rates make $180 worth

Windsurf's live model table exposes the rates used after the included quota. Model choice changes the invoice far more than the length of the plan name.

Use one normalized workload to compare them: ten agent sessions, each with 100,000 input tokens and 10,000 output tokens. Across all ten sessions, that is 1 million input tokens and 100,000 output tokens. This is analysis, not a claim that every coding session has that shape. A small repository can use less. A long-running agent with a large shared timeline can use much more.

At current list rates:

• Adaptive costs $0.50 per million input tokens and $2 per million output tokens. The normalized workload costs $0.70.
• Claude Sonnet 5 Medium costs $2 per million input and $10 per million output. The same workload costs $3.00.
• Claude Opus 5 Medium costs $5 per million input and $25 per million output. The same workload costs $7.50.

Prompt caching can reduce repeated input. If all 1 million input tokens qualify as cache reads, the same workloads fall to $0.30 on Adaptive, $1.20 on Sonnet 5 Medium, and $3.00 on Opus 5 Medium. Real sessions mix new input, cached context, and output, so the usage meter is the source of truth.

Now apply the Max premium. The move from Pro to Max costs $180/month. At extra-usage list rates, $180 buys about 257 Adaptive workloads, 60 Sonnet 5 Medium workloads, or 24 Opus 5 Medium workloads of the normalized ten-session size.

This is not proof that Pro includes those quantities. The quota remains undisclosed. It is the opportunity cost of upgrading: Max has to deliver more value than the same $180 spent on measured overage, or remove interruptions that are worth at least $180 to the operator.

For a funded founder, one blocked release day may clear that threshold immediately. For a solo technical builder, 24 large Opus workload packs may be more than the monthly burst requirement. For a mid-market engineering lead, Max is the wrong organizational answer because it remains a personal plan; Teams or Enterprise handles governance.

Teams pricing is a minimum, not an $80 base fee

Windsurf Teams costs at least $80/month, but the detailed billing documentation defines that $80 as a minimum, not a permanent platform fee added on top of every seat. Full-seat charges satisfy the minimum.

This point is easy to misread on the pricing card. The card says “$80/month for team plan + $40/mo per full dev seat.” The detailed self-serve combinations resolve the billing behavior:

• 0 full seats: $80 total, all $80 becomes shared on-demand credits.
• 1 full seat: $80 total, with $40 funding the seat and $40 becoming shared on-demand credits.
• 2 full seats: $80 total, with both $40 seats satisfying the minimum and no prepaid credits included.
• 3 full seats: $120 total.
• 5 full seats: $200 total.
• 10 full seats: $400 total.

Once a team has two or more full seats, the formula is simply the number of full seats multiplied by $40. Each full seat includes a Pro-equivalent daily and weekly quota plus access to Devin Desktop. Paid usage beyond those allowances comes from the team's shared on-demand balance.

Flex seats are free and unlimited, but they are not free developers. A flex seat has no included quota and no Devin Desktop access. Every task draws from shared on-demand credits. Flex works for an occasional approver, product manager, or teammate who needs ad-hoc access. Giving a regular developer a flex seat turns a predictable $40 line into uncapped shared consumption.

The comparison with separate Pro subscriptions shows what Teams sells. Five individual Pro plans cost $100/month. Five Teams full seats cost $200. Ten Pro plans cost $200; ten Teams seats cost $400. At two or more seats, Teams is exactly 2× the individual Pro seat price.

That premium buys central billing, shared collaboration, an admin dashboard with analytics, and priority support. It does not buy a larger included quota per full developer; the documentation calls each full-seat quota equivalent to Pro. New Teams plans also do not include SSO. SAML/OIDC SSO is an Enterprise feature.

A five-person startup should therefore ask one operational question: are administration, shared usage, collaboration, and support worth $100/month more than five independent Pro accounts? If the answer is no, separate Pro accounts are cheaper. If access control, offboarding, consolidated billing, and shared agents matter, the Teams premium is a normal governance cost.

Hidden costs, annual billing, and the March 2026 change

Windsurf's largest hidden cost is not a surcharge. It is the combination of unpublished included quotas, API-priced continuation, and automatic reload. A $20 plan can remain $20, but the public page cannot tell you how much frontier-model work fits before the meter starts.

Overage follows tokens, model choice, and speed

Extra usage is charged at API list prices for the selected model. A short instruction can still be expensive when the agent sends a large codebase, system prompt, and shared conversation timeline as context. Output and reasoning add their own tokens. Faster or higher-priority configurations cost more.

The spend controls are useful only when configured. On-demand credits can reload automatically. Windsurf lets admins set reload thresholds and default per-session spending limits, but an enabled reload without a review owner can turn quota exhaustion into a quiet recurring charge. Purchased credits do roll over and never expire, so a conservative manual top-up has little expiry downside.

Daily quotas changed the economics of bursty work

In March 2026, Windsurf replaced the old monthly prompt-credit pool for self-serve customers with daily and weekly quotas. A monthly pool let a developer spend most of the allowance on one large project. The new Pro shape paces that work with both clocks. Max removes the daily cap but keeps a weekly allowance.

This is the business consequence of the pricing change: the same monthly subscription can support steady daily work and still block a valuable one-day sprint. A founder planning a migration should price interruption, not average prompts per month.

Existing subscribers received grandfathering. A legacy Pro subscriber at $15/month keeps that price indefinitely. A legacy Teams subscriber at $30/month per developer seat also keeps the old price indefinitely. Existing paid subscribers received a seven-day extension during the migration.

Current self-serve pricing is monthly only

The public Free, Pro, Max, and Teams pages list monthly prices and no current annual discount for new self-serve buyers. The documentation still discusses existing annual Pro and Teams subscriptions because they predate the quota migration. Those subscribers received the seven-day extension and may request a refund for all remaining months if they cancel after the change.

For a new buyer, there is no public annual-lock discount to evaluate. For a grandfathered annual buyer, the migration refund right matters more than a generic cancellation assumption.

The refund promise is narrow

Windsurf says a customer who subscribes without receiving a trial can request a refund before using the subscription if they change their mind. That is not a blanket refund guarantee after paid usage. Paid individual plans can otherwise be canceled at any time and remain active through the current billing period before reverting to Free.

Converted legacy add-on balances have a separate rule: they can be refunded at any time through support. Do not blend that migration policy with the ordinary new-subscription statement.

“1,000 credits” is now a legacy question

Current self-serve Windsurf no longer sells the old monthly prompt-credit allowance. The March migration converted unused add-on credits into dollar-denominated extra usage at the original $0.04 per credit rate. Every 250 add-on credits became $10.

The official usage documentation still lists $120 for 1,000 pooled credits for accounts on the legacy credit-based enterprise model. That answers the pulled query, but it should not be used to forecast a new Free, Pro, Max, or Teams account. New self-serve overage follows model API prices instead.

One vendor contradiction needs a manual check

The live pricing page says Pro includes free use of SWE-1.7. The live quota and model documentation still describe an SWE-1.7 free-preview end date of 8 August 2026, five days before this verification. The pricing-page entitlement is the current public sales claim, but the stale date makes the model selector the only safe confirmation before assigning free SWE-1.7 a budget value.

That discrepancy is also a useful buying rule. Never justify Max, or any overage forecast, with a temporary zero-priced model until the current selector confirms it.

Windsurf vs Cursor, GitHub Copilot, and Claude Code on price

Windsurf Pro is tied with Cursor Pro and monthly Claude Pro at $20, while GitHub Copilot Pro starts at $10. The dollar tie hides four different products: an AI editor joined to Devin agents, another AI editor with disclosed usage pools, a completion-first GitHub assistant, and a terminal-first agent included in a broader Claude subscription.

Cursor: the more auditable $20 editor

Cursor Pro costs $20/month and publishes at least $20 of third-party model usage plus generous use of Cursor's own model pool. Windsurf Pro costs the same but does not publish a dollar or token value for its included quota.

Cursor Pro Plus costs $60/month and includes $70 of third-party model usage. Ultra costs $200 and includes $400. After the included pool, on-demand usage continues at the same API rates without downgrading request quality or speed. Cursor's own documentation says daily Agent users typically consume $60 to $100 of total usage per month, which gives a buyer a planning range even before personal data exists.

That transparency makes Cursor the cleaner budget when your finance question is “how much frontier-model value is bundled?” Windsurf can still win on product fit, especially if its editor, Devin Cloud agents, and workflow feel better for the codebase. Price alone does not prove that because the Windsurf allowance is hidden.

For the capability decision rather than the invoice, use the full Windsurf vs Cursor comparison. The pricing rule is narrower: tied at $20, Cursor discloses the pool and Windsurf asks you to discover it through usage.

GitHub Copilot: the cheapest paid seat

GitHub Copilot Pro costs $10/user/month, includes unlimited code completion, and carries $15 in total monthly AI credits across agent mode, code review, cloud agents, CLI, and other credit-consuming features.

Copilot Free costs $0 and includes 2,000 completions per month. Pro+ costs $39 with $70 in monthly AI credits. Max costs $100 with $200 in credits. Additional credits cost $0.01 each, so a $10 overage budget buys 1,000 credits.

Copilot is the budget pick for a developer who wants excellent completion, agent access inside an existing editor, and GitHub-native workflows without switching to Windsurf. It is not a direct replacement for Windsurf's integrated editor and Devin environment. That is why the $10 difference is meaningful for a completion-first user and less meaningful for someone choosing an agentic workspace.

Verified students also have access to GitHub Copilot Student. Windsurf lists no public student plan or discount, so Copilot is the clearer zero-dollar option for an eligible student.

Claude Code: $20 for a terminal agent and the rest of Claude

Claude Pro costs $20/month, or $17/month equivalent with $200 billed up front annually, and includes Claude Code. The subscription also covers Claude across web, desktop, and mobile, with all of those surfaces drawing from the same usage pool.

Claude Max starts at $100/month and offers 5× or 20× Pro usage options. Paid plans can turn on usage credits at standard API rates after the allowance. Claude therefore offers the strongest annual discount of this shortlist, but its shared pool means a long Claude Code session also consumes capacity used by chat and other Claude products.

Choose Claude Code when the terminal is the preferred control surface and the broader Claude subscription has value. Choose Windsurf when the editor, inline completion, and graphical agent workflow are part of the purchase. The Claude Code vs Cursor guide explains the terminal-versus-IDE tradeoff; the same workflow distinction applies here.

The normalized buyer rule is blunt:

• Windsurf Pro at $20: best when you want Windsurf plus Devin agents and will measure the undisclosed quota yourself.
• Cursor Pro at $20: best when you want an AI editor with a disclosed $20 third-party usage pool.
• GitHub Copilot Pro at $10: best when completion and GitHub integration matter more than changing editors.
• Claude Pro at $20 monthly or $17 annual: best when Claude Code and the rest of Claude share one useful subscription.

Which Windsurf plan should you choose?

Choose Free for occasional coding, evaluation, or a workflow dominated by unlimited Tab completions and inline edits. You never need to pay if agent resets do not interrupt anything valuable.

Choose Pro for a solo founder, operator, or developer who uses Windsurf in paid work. It is the only plan with a reasonable default price and a continuation path after quota. Add a controlled on-demand balance before assuming Max is necessary.

Choose Max only after Pro data shows a repeated daily-cap problem or more than $180/month of combined overage and interruption cost. Its no-daily-cap shape is useful for bursty work. Its undisclosed weekly allowance makes a speculative upgrade impossible to defend.

Choose Teams when at least two developers need one organization, shared administration, usage analytics, collaboration, and support. Budget $40 per full seat once the $80 minimum is satisfied. Use flex seats for genuinely occasional members, never as a hidden substitute for regular full seats.

Choose Enterprise when SSO, centralized enterprise controls, dedicated deployment, procurement terms, or account support are requirements. Enterprise usage is contracted in Agent Compute Units at the rate in the order form, so there is no public list-price comparison to make.

The Monday move is to buy data before capacity. A seven-day baseline is enough to expose whether the daily limit, weekly limit, or model selection is the actual constraint.

1. 1

   Start on Pro, not Max

   Put one regular developer on the $20 Pro plan. Keep the work representative: normal repository, normal model choices, and the real mix of Tab, local agent, CLI, and cloud-agent tasks.

2. 2

   Cap on-demand usage

   Open Settings, then Usage, and set a small reload threshold plus a default session spending limit. Manual top-ups are also reasonable because purchased credits roll over and never expire.

3. 3

   Record both reset clocks

   For seven days, note when the daily allowance interrupts work, when the weekly allowance becomes tight, which model was active, and how much paid usage continued the task. Do not average away the day that mattered.

4. 4

   Approve the consequence

   Upgrade to Max only if the measured overage plus lost time exceeds $180/month or the no-daily-cap workflow is worth that amount. Move to Teams only when shared administration earns the 2× seat premium.

Windsurf pricing FAQ

Is Windsurf still free?

Yes. Windsurf Free costs $0 and includes unlimited Tab completions and inline edits plus a light agent quota. Free users have limited model access, currently listed as SWE-1.6 only, and must wait for the next daily or weekly reset after the agent allowance is exhausted.

How much does Windsurf cost per month?

Free costs $0, Pro costs $20/month, and Max costs $200/month. Teams has an $80 monthly minimum and costs $40 per full seat once two seats satisfy that minimum. Enterprise pricing is custom.

How much is 1,000 Windsurf credits?

The current self-serve plans no longer use the old monthly prompt-credit allowance. Official documentation lists $120 for 1,000 pooled credits on legacy credit-based enterprise accounts. New self-serve overage is billed at model API list prices through on-demand credits.

Does Windsurf offer a student discount?

No student plan or student discount is listed on Windsurf's public pricing page as of 13 August 2026. Eligible verified students can get GitHub Copilot Student, making it the direct free paid-tier alternative in this category.

What is Windsurf's refund policy?

Windsurf says a new subscriber can request a refund before using the subscription if they change their mind. It also allows refunds of unspent extra-usage balances converted from legacy add-on credits, and documented migration terms let affected annual subscribers request refunds for remaining months. These are specific cases, not a blanket refund guarantee after usage.

Does Windsurf offer annual billing?

The current public self-serve page lists monthly prices only and publishes no annual discount for new Free, Pro, Max, or Teams buyers. Legacy annual Pro and Teams subscriptions are still documented because they existed before the March 2026 quota migration.

Did Windsurf pricing change in 2026?

Yes. In March 2026, self-serve billing moved from a monthly prompt-credit pool to daily and weekly quotas. New Pro is $20/month and Max is $200/month. Existing $15 Pro subscribers and $30-per-developer Teams subscribers kept their grandfathered prices indefinitely.

Which is cheaper, Cursor or Windsurf?

Windsurf Pro and Cursor Pro both cost $20/month. Cursor publishes at least $20 of included third-party model usage, while Windsurf does not publish the value of its quota. The cheaper one at real usage cannot be known until Windsurf's meter supplies the missing allowance data.

Who should buy Windsurf Max?

Buy Max only if Pro's daily cap repeatedly blocks valuable work or if measured on-demand usage plus interruption cost exceeds the $180 monthly upgrade. Max removes the daily cap and raises the weekly quota, but Windsurf does not disclose the exact allowance.

Get the AI Business Workflow Audit Checklist and the operator newsletter.