Windsurf
Windsurf pricing now runs on a daily and weekly token quota rather than the flow action credits it launched with, and the product itself is called Devin Desktop.
Updated on:
Windsurf pricing: the credits are gone and so is the brand
Windsurf pricing now runs on a daily and weekly token quota rather than the flow action credits it launched with, and the product itself is called Devin Desktop. Cognition bought Windsurf in July 2025, repriced it in March 2026, and renamed it that June. Visit windsurf.com/pricing today and a 308 redirect lands you on devin.ai/pricing. The useful question is whether Cognition now runs one meter or two.
Key takeaways
Plans run Free, Pro at $20, Max at $200 and Teams at an $80 monthly minimum, full seats $40 each.
Credits died on 19 March 2026. Your plan buys a token budget that refreshes daily and weekly.
Cognition publishes the conversion it uses between the two units: 12.5 legacy prompt credits equal one Agent Compute Unit.
Pro subscribers from before the change keep $15 a month indefinitely, and Teams seats keep $30.
Windsurf pricing in 2026
Plan | Price | Members | Included usage |
|---|---|---|---|
Free | $0 | 1 | Light quota, limited model choice |
Pro | $20/mo | 1 | Daily and weekly quota, full model choice |
Max | $200/mo | 1 | Significantly higher quotas |
Teams | $80/mo minimum | Up to 200 | Full seats $40/mo each, with their own quota |
Enterprise | Custom | Unlimited | Contracted ACU volume |
What Windsurf actually meters
Windsurf meters tokens, and the rate depends entirely on which model you pick. Since March 2026 your plan carries a budget measured in model tokens, and the docs publish input, output, cache write and cache read rates for every model. Short requests with a few files in context cost less than long ones over a large codebase, which is the behaviour the old credit system could not price.
Cognition's own SWE models carry the discount. On paid self-serve plans SWE-2 and SWE-1.7 are listed at zero cost per million tokens, free through 15 October 2026, after which SWE-2 reverts to list at $3.00 input and $15.00 output per million. SWE-1.6 still bills at $0.50 and $2.50. Third party models cost what they cost: Claude Opus 4.5 is $5.00 input and $25.00 output.
Enterprise runs a second unit on the same meter. Those contracts bill in Agent Compute Units, and for local agents such as Cascade and the CLI an ACU is nothing more than tokens converted at those same per-model rates. The models page carries the constant it divides by: 12.5 legacy prompt credits to one ACU. A model that cost 6 prompt credits a message therefore costs 6 divided by 12.5, or 0.48 ACU.
So Cognition runs one meter in two denominations. Both resolve to the same per-model token rates. The Devin entry in this index describes an ACU as agent effort, which still holds for cloud agents where virtual machines and infrastructure enter the mix. For anything on your laptop it is tokens.
What happens when you hit the limit
Paid plans keep working and bill the difference. Past your included quota, Pro, Max and Teams users buy extra usage charged at the model's API list price, and Cognition states plainly that quota limits never cap extra usage. Free users wait for the next daily or weekly reset.
The reset schedule has one detail worth knowing. Your daily quota is deliberately set higher than one seventh of the weekly figure, so somebody working weekends can still spend the full week's allowance. Priority and speed variants, anything labelled Fast or Lightning, consume more. Legacy enterprise accounts still on credits refill in multiples of $120 once the balance drops below 15 credits, against a default monthly cap of $160.
How Windsurf's pricing has changed across all these years
Windsurf has repriced five times in two years and blogged every one of them.
Date | Milestone | Source |
|---|---|---|
2 Jun 2026 | Windsurf renamed Devin Desktop. windsurf.com/pricing starts redirecting to devin.ai/pricing | devin.ai/blog and docs.devin.ai |
14 Apr 2026 | Devin's self-serve plans move to the same Free, Pro $20, Max $200, Teams $80 lineup | devin.ai/blog and docs.devin.ai |
18 Mar 2026 | Credits replaced by daily and weekly token quotas, effective 19 March. Max launched at $200. Pro rises to $20 and Teams to $40 a seat, with existing subscribers grandfathered at $15 and $30 | devin.ai/blog and docs.devin.ai |
14 Jul 2025 | Cognition acquires Windsurf, three days after Google licensed the technology and hired the founders | devin.ai/blog and docs.devin.ai |
28 Apr 2025 | Free tier raised from 5 premium prompt credits a month to 25 | devin.ai/blog and docs.devin.ai |
21 Apr 2025 | Flow action credits eliminated. You pay per user prompt however many steps Cascade takes. Pro $15 for 500 credits, Teams $30 a seat, Enterprise $60 a seat | devin.ai/blog and docs.devin.ai |
6 Dec 2024 | Cascade priced for the first time, after the launch step limiter was scrapped mid-flight | devin.ai/blog and docs.devin.ai |
Source: devin.ai/blog and docs.devin.ai, read 9 October 2026.
Customer
Sentiment Highlights
"the $15/mo subscription was the best value on Earth paired with weekly deals reducing credits for premium models. Now it's barely any messages for paid models, completely watered down"
Former Windsurf subscriber on Devin Desktop, Hacker News, July 2026
"I used to use windsurf as my main editor until they changed their pricing model"
Grandfathered Windsurf user, Hacker News, September 2026
Explore other providers
Devin
Developer Tool
Devin pricing meters an Agent Compute Unit, and the useful thing about it is how precisely Cognition documents what does and does not consume one.
Cursor
Developer Tool
Cursor pricing combines a flat monthly subscription with token usage billed at each model's API rate.
GitHub Copilot
Developer Tool
GitHub Copilot pricing stopped counting requests in June 2026 and started counting money.
How much does Windsurf cost per month?
Does Windsurf still use credits?
What happens when you run out of Windsurf quota?
Is Windsurf the same as Devin?
