Devin Pricing: Windsurf Pricing, Devin AI Pricing, Teams Seats and ACU Cost
Most Devin pricing guides still quote a $20 Core plan and a $500 Team plan. Cognition retired both in April 2026, and since then windsurf.com/pricing has redirected to the same Devin price list. This page lists what Devin costs today in US dollars, works out what a team actually pays, and shows the only public per-ACU rates, which sit on cloud marketplace listings rather than on devin.ai.
No agents in this category in the preview.
See the full catalog →What it does
What it does not do
Interactive preview · no account needed
Not connectable yet
This agent cannot be deployed today: we have not shipped the connector it needs. There is nothing to try out here yet, and the full listing says the same.
Open the full listing →In short
Devin has five plans. Free is $0 with a light agent quota. Pro is $20 a month and Max is $200 a month, each for one user with up to 10 concurrent sessions. Teams costs $80 a month as a minimum, plus $40 a month for each full developer seat, with unlimited free flex seats that draw from a shared credit pool. Enterprise is quoted by sales and billed in ACUs (Agent Compute Units). Self-serve plans no longer sell ACUs: usage past your included quota is billed in dollars at API pricing. The only public per-ACU prices are on cloud marketplaces, $4.00 per additional ACU on AWS and $3.00 per ACU on the Microsoft Marketplace. Windsurf pricing is now Devin pricing, because windsurf.com/pricing redirects to devin.ai/pricing.
At a glance
Devin plans and US prices, read from devin.ai/pricing and docs.devin.ai on September 16, 2026
| Plan | US price | Users | Concurrent sessions | What it adds |
|---|---|---|---|---|
| Free | $0 | 1 | Not included | Light agent quota, limited models, unlimited Tab completions |
| Pro | $20 per month | 1 | Up to 10 | Frontier models from OpenAI, Anthropic and Google, cloud agents, Devin API, extra usage at API pricing |
| Max | $200 per month | 1 | Up to 10 | Everything in Pro with significantly higher quotas, weekly allowance only |
| Teams | $80 per month minimum, plus $40 per full seat | Up to 200 self-serve | Unlimited | Shared credit pool, free flex seats, admin analytics, centralized billing |
| Enterprise | Quoted by sales, billed in ACUs | Unlimited | Unlimited | SAML or OIDC SSO, VPC deployment, per-session ACU caps, dedicated account team |
The challenge
The price list is short, but the pages that rank for it are out of date. Cognition announced on April 14, 2026 that it was "retiring the old Core and Team plans," replacing "the old $500/month Team entry point," and moving self-serve overage from ACUs to dollars. Guides written before that still tell buyers to budget $2.25 per ACU on a pay-as-you-go Core plan that no longer exists. The second source of confusion is the Windsurf merger: a developer who searches for Windsurf pricing lands on the Devin page, sees Devin Desktop, and cannot tell whether it is the same product or a different bill. It is the same price list.
How Agentmarketplace handles it
Start from seats, not from the model. If one developer uses Devin, compare Pro at $20 with Max at $200 and ignore Teams. If three or more people need it, Teams is the default, and the real question is how many of them need a full $40 seat with their own quota and Devin Desktop, and how many can be free flex seats pulling from the shared pool. Next, decide whether anyone will run long cloud sessions, because that is what burns quota and triggers dollar overage. Finally, if your company needs SSO, VPC deployment or per-session spend caps, you are on Enterprise and the price comes from an order form. And if the goal is one well-defined job, such as triaging bugs or reviewing pull requests, a ready-made coding agent from a catalog can be cheaper than a general agent seat.
How much does Devin cost?
Devin costs $0, $20 or $200 a month for one person, and a team pays at least $80 a month. Those are the four self-serve prices on devin.ai/pricing, and Enterprise sits above them with a contract price. There is no annual discount on the page: every self-serve price is monthly.
What changes between the plans is quota, not features. Pro and Max carry the same feature list, including cloud agents, the Devin API, DeepWiki, Ask Devin and integrations with Slack, Linear, Jira and GitHub. Max simply has "significantly higher quotas." Free is the plan that is missing things: no premium models, no concurrent cloud sessions, and none of the integrations.
| Feature | Free | Pro, $20 | Max, $200 | Teams | Enterprise |
|---|---|---|---|---|---|
| Premium frontier models | No | Yes | Yes | Yes | Yes |
| Cloud agents and Devin API | No | Yes | Yes | Yes | Yes |
| Slack, Teams, Linear and Jira | No | Yes | Yes | Yes | Yes |
| GitHub, GitLab and Bitbucket | No | Yes | Yes | Yes | Yes |
| Sharing and collaboration | No | No | No | Yes | Yes |
| SAML or OIDC SSO | No | No | No | No | Yes |
| Deploy in your own VPC | No | No | No | No | Yes |
| Usage allowance | Light | Daily and weekly | Weekly only, higher | Per full seat plus shared pool | ACUs per order form |
Cognition does not publish quota sizes as numbers. Its pricing FAQ says each paid plan has "a usage allowance that refreshes automatically on a daily and weekly basis" and that "for the majority of users, this quota will be enough to fully cover all agent usage." Past that point, "you can purchase extra usage which is consumed at API pricing." Choosing smaller models, such as Haiku or GPT 5.2 Mini, stretches the allowance further.
One promotional line is worth dating: SWE-2, Cognition's own coding model, is "Free in Devin Desktop and CLI through October 10, 2026." After that date, SWE-2 use in those tools may start drawing on quota, so a team that leans on it should re-check the plan page in October.
Windsurf pricing is now Devin pricing
If you search for Windsurf pricing, you end up on Devin's price list. On September 16, 2026, windsurf.com/pricing answered with an HTTP 308 permanent redirect to devin.ai/pricing, and the page it lands on is titled "Plans and Pricing | Devin." Cognition bought Windsurf in 2025, and the editor is now sold as Devin Desktop.
For a buyer, that means there is one bill, not two. The same Free, Pro, Max, Teams and Enterprise plans cover the desktop editor, the CLI and Devin's cloud agents. Tab completions are unlimited on every plan, including Free, which is what most former Windsurf users relied on day to day.
| If you were on | You are now on | What to check |
|---|---|---|
| Devin Core (pay as you go) | Devin Free | Cognition says Free "does not preserve the old no-minimum pay-as-you-go structure." Ongoing paid usage needs Pro, Max or Teams |
| Devin Team ($500 a month) | Devin Teams | The new entry point is $80 a month, so small teams usually pay less |
| A Windsurf individual plan | Devin Free, Pro or Max | Devin Desktop access on Teams requires a full seat, not a flex seat |
| A Windsurf team plan | Devin Teams or Enterprise | SSO and centralized multi-org management are Enterprise only |
The practical consequence for teams: a flex seat on Devin Teams is free, but Cognition's docs say flex seats "do not include Devin Desktop access." Anyone who wants the editor needs a $40 full seat. That single rule decides most Teams bills, which is why the next section works through it.
Devin Teams pricing: the $80 minimum and the $40 full seat
Devin Teams costs $80 a month or $40 per full seat, whichever is higher. Cognition's billing docs spell out the math: with no full seats, the $80 becomes on-demand credits for the team; with one full seat, you pay $40 for the seat plus $40 of credits; from two full seats up, you pay $40 per seat and the minimum is already met.
| Full developer seats | Monthly bill | Annual bill | What the minimum becomes |
|---|---|---|---|
| 0 | $80 | $960 | $80 of shared on-demand credits |
| 1 | $80 | $960 | $40 seat plus $40 of credits |
| 2 | $80 | $960 | Two seats, no credits |
| 5 | $200 | $2,400 | Five seats |
| 10 | $400 | $4,800 | Ten seats |
| 25 | $1,000 | $12,000 | Twenty-five seats |
| 50 | $2,000 | $24,000 | Fifty seats |
The annual column is simply twelve months at the published monthly rate; Cognition does not publish an annual discount. Each full seat carries "a daily and weekly usage quota equivalent to the Pro plan" plus Devin Desktop. Flex seats cost nothing and "draw entirely from the team's shared pool of on-demand credits," which makes them the right choice for product managers, reviewers and occasional users who kick off a session a few times a week.
Two limits matter for larger companies. The pricing page says "Unlimited team members," but the self-serve billing docs cap Teams at 200 members and say larger teams "should contact sales about Enterprise." And on-demand credits "roll over month-to-month," and "purchased credits never expire," so over-buying credits is not wasted money the way it is on some annual-term platforms.
Devin Teams against the old Team plan
Under the retired Team plan, the entry point was $500 a month. A five-person team now pays $200 with five full seats, or $80 if only two people need the editor and the rest use flex seats. The trade is that the old plan bundled a block of ACUs, while the new one bundles per-seat quota and bills overage in dollars.
What is an ACU in Devin, and what does one cost?
An ACU, or Agent Compute Unit, is Cognition's measure of how much work Devin did. The docs define it as "the amount of agent effort required to complete a given task," and for cloud agents "a mix of tokens, compute, VMs, and other infrastructure costs." Virtual machine time and bandwidth are "typically a small fraction of total usage." Windows sessions use about 9% more than Linux.
ACUs still matter, but only on Enterprise. Since April 2026, self-serve usage counts against quota and anything above it "will be priced and billed in dollars rather than ACUs," while "Enterprise billing will continue to use ACUs" at the rate set in each order form. The old $2.25 and $2.00 per ACU figures that many guides still quote belonged to the retired Core and Team plans and do not appear on any current Cognition page.
| Where Devin is listed | Published ACU price | Status on September 16, 2026 |
|---|---|---|
| AWS Marketplace, "Devin by Cognition" | $4.00 per additional ACU | 12 month contract; base contract price is set by private offer |
| Microsoft Marketplace, "Devin by Cognition" | $3,000 per month including 1,000 ACUs, then $3.00 per ACU | Listing shows a scheduled price change on December 1, 2026 to a placeholder value |
| Google Cloud Marketplace | Not published | Private offer only |
| devin.ai/pricing | Not published | Enterprise says "Let's talk" |
So the defensible public range for an Enterprise ACU is $3 to $4 at list, with the real number negotiated. Note that the Microsoft listing's $3,000 plan is being replaced by what looks like a placeholder from December 1, 2026, so treat it as a dated reference point rather than something you can still buy after that date.
How many ACUs does a Devin session use?
Cognition publishes size bands rather than per-hour rates. For self-serve accounts a session is XS at up to 2 ACUs, S up to 5, M up to 10, L up to 20 and XL above 20, and sessions classed L or XL are "flagged as unhealthy." For Enterprise, the thresholds are "scaled by a factor of 10," so an Enterprise XS session is up to 20 ACUs. That scaling means an Enterprise ACU is a smaller unit than the old self-serve ACU, which is one more reason not to compare old self-serve rates with Enterprise quotes. Other published reference points: a medium-effort DeepWiki costs about 5 to 10 ACUs per wiki, a high-effort one 20 to 40, and Fast Mode uses 4 times the ACUs per session for roughly 2 times faster responses.
Devin also "sleeps automatically after 30 minutes of inactivity by default" and uses nothing while asleep. Enterprise admins can set that timeout between 5 and 120 minutes, plus hard ACU caps per session, which are the two controls that stop an idle or looping session from running up the bill.
Devin pricing compared with Cursor, GitHub Copilot, Claude Code and Codex
Devin's $20 and $200 price points match the rest of the market almost exactly, so the comparison comes down to team pricing and what each vendor counts. All prices below are US list prices read from each vendor's own pricing page on September 16, 2026.
| Tool | Individual plans | Team price | How usage is counted |
|---|---|---|---|
| Devin (includes Windsurf) | $0, $20, $200 a month | $80 minimum, $40 per full seat, free flex seats | Daily and weekly quota, dollar overage at API pricing |
| Cursor | Pro $20, Pro+ $60, Ultra $200 ($16, $48, $160 yearly) | $40 Standard or $120 Premium per user ($32 or $96 yearly) | Included model usage, on-demand usage billed in arrears |
| GitHub Copilot | Pro $10, Pro+ $39, Max $100 | Business $19, Enterprise $39 per seat | GitHub AI Credits at $0.01 each, pooled across the org |
| Claude Code (Claude plans) | Pro $20 ($17 yearly), Max $100 or $200 | Team $25 Standard or $125 Premium ($20 or $100 yearly) | Per-seat usage limits; Enterprise usage at API rates |
| OpenAI Codex (ChatGPT plans) | Plus $20, Pro from $100 | Business $25 per user ($20 yearly) | Plan usage limits plus credits |
| Factory | Pro $20, Plus $100, Max $200 | $60 per team plus $40 per seat, up to 10 seats | Plan usage tiers |
| Replit | Core $20 ($18 yearly), Pro from $100 | Pro includes up to 15 collaborators | Monthly credits, then effort-based pricing |
Three things stand out. First, Devin is the only one here with free team seats, which makes it cheap for teams where most members only review or launch work. Second, Devin and Factory both use a team base fee plus $40 per seat, while Cursor and Claude charge per seat with no base fee. Third, Devin publishes no annual discount at all, while Cursor, Claude and ChatGPT all do, so a yearly commitment narrows or reverses Devin's price advantage for heavy individual users.
Is Devin cheaper than Cursor for a team?
For a ten-person team where everyone needs the editor, Devin Teams is $400 a month and Cursor Teams Standard is $400 on monthly billing or $320 on yearly billing. So Cursor is cheaper once you commit for a year. If only four of the ten need an editor seat, Devin drops to $160 a month because the other six can be flex seats, which Cursor has no equivalent for.
Devin Enterprise pricing and how to buy it
Devin Enterprise has no public price. The docs say Enterprise customers are "billed in Agent Compute Units (ACUs) at the rate set in their order form," and those ACUs "are distinct from self-serve quota and on-demand credits." What Enterprise adds is control: per-organization ACU limits, per-user usage policies, hard ACU caps per session, SAML or OIDC SSO, teamspace isolation, centralized management of multiple Devin organizations, and the option to deploy in your own virtual private cloud.
US companies that already buy through a cloud provider have a procurement shortcut. Devin is listed on AWS Marketplace, and on September 15, 2026 AWS announced a multi-year collaboration with Cognition, noting that joint customers "can already purchase Devin through AWS Marketplace." Buying there lets the spend count against an existing AWS commitment and skips a new vendor onboarding. A separate FedRAMP listing for Devin Desktop exists for public sector buyers, also on custom pricing.
Before you sign, get three numbers in writing: the price per ACU, the ACU volume included in the contract term, and what happens to unused ACUs at renewal. Self-serve credits never expire, but that promise is written for self-serve, and an Enterprise order form can say something different.
Which Devin plan should you buy?
Buy Pro if you are one developer who wants cloud agents and the Devin API, and move to Max only after you hit the weekly limit more than once a month. The jump from $20 to $200 buys quota, nothing else. Buy Teams as soon as two or more people share work, and give full seats only to people who need Devin Desktop or their own quota. Everyone else can be a free flex seat. Buy Enterprise when security review requires SSO or VPC deployment, or when finance needs per-session spend caps.
A general coding agent is not always the right purchase. Many teams want one repeatable job done well, such as reviewing every pull request, triaging new issues or keeping dependencies current. A focused agent built for that job is often cheaper to run and easier to budget than an open-ended agent seat, and you can compare them side by side in our AI agent catalog. For how agent vendors meter usage more generally, see our breakdown of AI agent pricing models.
Frequently asked
Devin pricing: your questions answered
How much does Devin cost?
Devin costs $0 on Free, $20 a month on Pro and $200 a month on Max, each for one user. Devin Teams costs $80 a month as a minimum, or $40 a month per full developer seat once you have two or more, with free flex seats. Enterprise is priced by Cognition sales and billed in ACUs.
Is Windsurf now Devin?
Yes, for pricing purposes. windsurf.com/pricing permanently redirects to devin.ai/pricing, and the editor is now sold as Devin Desktop under the same Free, Pro, Max, Teams and Enterprise plans. Cognition bought Windsurf in 2025, so there is one price list covering the editor, the CLI and Devin cloud agents.
What is an ACU in Devin?
An ACU, or Agent Compute Unit, measures the agent effort Devin spends on a task, combining model inference, compute, virtual machine time and bandwidth. Since April 2026 only Enterprise contracts are billed in ACUs. Self-serve plans use a daily and weekly quota and bill overage in dollars at API pricing.
How much does a Devin ACU cost?
Cognition does not publish an ACU price on devin.ai. The public reference points are cloud marketplace listings: $4.00 per additional ACU on AWS Marketplace, and $3.00 per ACU after 1,000 included ACUs on a $3,000 monthly Microsoft Marketplace plan, which shows a scheduled change on December 1, 2026. Enterprise rates are set in the order form.
How much does Devin Teams cost?
Devin Teams costs $80 a month minimum. With no full seats the $80 becomes shared credits, with one full seat it is $40 for the seat plus $40 of credits, and from two full seats it is $40 per seat. Ten full seats cost $400 a month. Flex seats are free but do not include Devin Desktop.
Is Devin still $500 a month?
No. Cognition retired the $500 a month Team plan and the pay-as-you-go Core plan in April 2026. Former Team customers moved to Devin Teams, which starts at $80 a month, and former Core customers moved to the Free plan. Guides quoting $500 or $2.25 per ACU predate that change.
Does Devin have a free plan?
Yes. Devin Free costs $0 for one user and includes a light agent quota, limited model access and unlimited Tab completions. It does not include premium models, concurrent cloud sessions, the Devin API or integrations with GitHub, Slack, Linear or Jira, which start on the $20 Pro plan.
Do Devin credits expire?
No, not on self-serve plans. Cognition states that on-demand credits roll over month to month and that purchased credits never expire, with optional auto-reload. Enterprise ACU terms are set in each order form, so confirm renewal and expiry terms in writing before signing an Enterprise contract.
Can I buy Devin through AWS Marketplace?
Yes. Devin by Cognition is listed on AWS Marketplace on a 12 month contract with custom pricing through a private offer and $4.00 per additional ACU. AWS announced a multi-year collaboration with Cognition on September 15, 2026. Buying there can count against an existing AWS spending commitment.
Is Devin cheaper than Cursor?
For individuals the prices match at $20 and $200 a month, but Cursor offers yearly billing at 20% off and Devin does not. For teams, Devin is cheaper when only some members need an editor, because flex seats are free. When everyone needs a seat, Cursor Teams at $32 a seat yearly beats Devin at $40.
Does Devin offer annual billing?
No annual price appears on devin.ai/pricing, and every self-serve plan is listed per month. That differs from Cursor, Claude and ChatGPT, which all publish yearly discounts. Enterprise contracts are negotiated, and the AWS Marketplace listing is sold as a 12 month contract.
Deploy your first Devin pricing agent today
Browse the marketplace, deploy in one click, and run on any model. No lock-in.