Devin Pricing (2026): $20 Is the Entry, Not the Budget

Devin Pro costs $20 a month, plus any extra usage at API pricing. For ongoing work after the scheduled SWE-2 offer, budget the subscription and measured usage: the public page gives no numeric agent quota or guaranteed SWE-2-specific rate after the promotion.

Devin lists the current Free, Pro, Max, Teams, and Enterprise tiers on its official pricing page. The plan card gives you the entry price. The detailed billing page explains seat types, quota resets, and credits that determine the bill.

As of October 17, 2026, Devin's pricing page still lists free SWE-2 in Devin Desktop and CLI through October 16, 2026, its scheduled end date. It does not state a replacement offer or a SWE-2-specific rate after that deadline. Prices and billing rules below were re-verified on October 17.

Since September 10, 2026, SWE-2, Cognition's coding model, has been available in Devin Desktop and CLI. Cognition explicitly calls Devin Desktop the new name for Windsurf, the editor; Devin CLI is its terminal coding agent. The listed offer names Desktop and CLI, with no promise of free Devin Cloud work. For the editor and legacy-plan context, see Windsurf pricing.

Devin pricing at a glance

Pro is the sensible solo default at $20 per month, while Teams advertises an $80 monthly plan fee plus $40 per full developer seat. Max costs ten times as much as Pro, yet Devin does not publish the numeric quota behind either plan. That makes Max a measured upgrade, not a safe default.

The table is the floor, not a forecast. By default, paid self-serve plans consume included quota first. Work beyond that allowance draws from prepaid on-demand credits at API pricing, and the public pages do not publish one fixed price per message or task.

The current pricing card and billing docs disagree on Teams. The card explicitly advertises $80 for the team plan plus $40 per full seat; the docs still describe an $80 minimum satisfied by seats and prepaid credits. The docs themselves direct readers to the pricing page for authoritative prices. Use the advertised formula for a new budget: two full seats are $160 and five are $280 before extra usage. Confirm the checkout terms before committing, rather than forecasting from the older minimum model.

What every Devin plan buys

The tiers separate light assistance, regular solo use, heavy solo use, shared team work, and contract deployment. The meaningful differences are who can use the account, when quota resets, whether usage can be pooled, and which controls surround the work.

Free: enough when interruptions are harmless

Devin Free costs $0 and can remain enough for occasional coding help. It includes a light quota for agent work, limited model availability, unlimited inline edits, and unlimited Tab completions. Devin's docs also list limited Devin usage plus access to Devin Review and DeepWiki.

The limit is deliberate: Devin does not publish a numeric free-agent allowance. You can evaluate the workflow, use inline edits, and keep accepting Tab completions, but you cannot build a production schedule around a promised number of autonomous tasks.

You may never need to pay if Devin is a secondary coding surface, your agent tasks are occasional, and waiting for the quota to refresh has no cost. Public GitHub pull-request reviews are also free at devinreview.com without a Devin account, so someone who wants only that public review path does not need a subscription.

Free stops being economical when a quota block delays paid work. At that point, the relevant comparison is not $0 versus $20. It is the value of the work waiting versus the cost of Pro.

Pro: the $20 working plan

Devin Pro costs $20 per month for one member and includes everything in Free, increased quotas, full model availability, and Devin Cloud access. Its daily and weekly allowance is shared across Devin sessions, Devin CLI, and Devin Desktop. The current Pro card names frontier models from OpenAI, Claude, Gemini, and SpaceXAI, plus leading open-source models, and says "Purchase extra usage at API pricing." The billing docs also list Slack, Linear, and MCP integrations. MCP, or Model Context Protocol, is the connector layer that lets an agent reach approved tools and data.

The pricing comparison lists up to 10 concurrent sessions. That is enough parallelism for most solo builders, but parallel work can consume a quota faster than one session at a time.

Pro's named weakness is disclosure. The allowance refreshes daily and weekly, yet Devin publishes neither quantity. Cost per message changes with the model, task size, complexity, and reasoning. The SWE-2 offer was scheduled to end on October 16, 2026; its continued appearance on the pricing page does not confirm an extension. It did not waive the $20 subscription or promise free cloud-agent work. The page gives the standard extra-usage rule but no replacement promotion or guaranteed SWE-2-specific token or task rate. Check the in-product rate and spending controls before continuing; on-demand credits can turn the $20 subscription into a variable line item.

Treat Pro as a one-month measurement plan. It proves whether autonomous delegation produces accepted work before you expose the card to larger quota purchases.

Max: a $200 answer to a measured problem

Devin Max costs $200 per month for one member and buys a significantly larger weekly allowance with no daily cap. The live pricing card still marks it NEW and describes significantly higher quotas. It includes everything in Pro and keeps the public limit at up to 10 concurrent sessions.

The absence of a daily cap matters for bursty work. A solo founder can run a concentrated migration or test-writing day without being stopped by Pro's daily allowance. The missing numeric weekly quota still prevents a clean task forecast before purchase.

Max is not automatically cheaper for a heavy user. It becomes the better budget only when its extra allowance displaces enough Pro overage, or when a daily stop would delay work worth more than the premium. Until the account's usage page proves one of those conditions, Pro plus a hard overage cap is the safer experiment.

Teams: an $80 plan fee plus $40 per full seat

Devin Teams advertises $80 per month for the team plan plus $40 per full developer seat, with up to 200 users. A full seat includes its own quota and Devin Desktop; the self-serve billing rules describe that allowance as Pro-equivalent. Those docs also describe free flex seats with no Desktop access, drawing only from shared on-demand credits.

Budget from the current pricing card:

• The team-plan fee is $80 per month before full-seat charges.
• One full seat adds $40, for $120 per month.
• Two full seats add $80, for $160 per month.
• Five full seats add $200, for $280 per month.

The older docs' seat-and-credit minimum examples do not match that advertised formula, so do not assume the $80 fee includes prepaid credits. Teams supports up to 200 members and unlimited concurrent sessions. It adds sharing, collaboration, centralized billing, an admin analytics dashboard, and priority support. On-demand credits are shared across the organization, so one active flex user can spend from the same pool another teammate expected to use.

The practical staffing rule is simple: give a full seat to a regular user and a flex seat to an occasional one. Consider a full seat when recurring credit use clears $40 a month and its included quota can cover that work, or when Devin Desktop is required. The team-plan fee remains a separate organization cost.

Enterprise: custom ACU economics

Devin Enterprise is a contract purchase, not a public seat price. Enterprise customers consume Agent Compute Units, or ACUs, at the rate written in their order form. An ACU is Devin's contract unit for the work the agent performs; current self-serve customers use quotas and dollar-denominated credits instead.

Enterprise adds SAML/OIDC single sign-on, centralized enterprise controls, dedicated deployment, dedicated account management, and the highest-priority support. It also lets enterprise admins set ACU limits by organization and inspect consumption across organizations.

Do not reuse an old public ACU rate in an Enterprise forecast. The current official rule is the order form, which makes any generic per-ACU number unreliable.

The break-even between Devin plans

The only defensible Pro-to-Max break-even is $180 of recurring monthly overage or equivalent blocked-work value. Max is $200 and Pro is $20, so the upgrade spends an extra $180 before it produces a single additional accepted result.

If Pro plus on-demand usage totals $120, Max still costs $80 more. If Pro plus recurring on-demand usage reaches $210, Max is cheaper by $10, assuming Max's larger allowance covers the same work. The second clause matters because Devin does not publish the two numeric quotas. Any precise task threshold would manufacture certainty the vendor does not provide.

The daily cap creates a second decision rule. Max can pay back below $180 of cash overage when Pro repeatedly blocks a valuable one-day sprint. That is a business calculation: attach a dollar value to the delayed release or engineering hour, then compare it with the premium. Do not call impatience a return on investment.

The flex-to-full-seat comparison starts at the $40 marginal seat fee. A flex seat costs $0 but uses shared on-demand credits. A full seat adds $40 and brings its own Pro-equivalent allowance plus Devin Desktop. If an occasional teammate regularly uses more than $40 of shared credits and the full-seat quota covers the same work, the seat can lower the direct usage bill. Full-seat overage can still add credits, and the $80 organization fee is separate.

Teams carries a collaboration premium over isolated Pro subscriptions. Two Pro accounts cost $40 in total, while the advertised Teams budget with two full seats is $160. Five Pro accounts cost $100, while Teams with five full seats is $280. Pro accounts cannot be shared, so that difference buys central billing, collaboration, pooled credits, analytics, and administration rather than more people at a lower seat price.

What one accepted coding outcome costs

Your useful unit is an accepted task, not a message, session, or ACU. An accepted task is a change you reviewed and kept: a merged fix, a passing test addition, a completed migration step, or another result that reached your quality bar.

At the subscription floor, Pro costs $4 per accepted task if it produces five in a month, $1 at 20 tasks, and $0.40 at 50. Max costs $10 per accepted task at 20 accepted tasks and $4 at 50. The advertised five-full-seat Teams budget is $280; if each developer accepts ten tasks, the team gets 50 accepted outcomes at a $5.60 plan-only cost each.

Those figures are budget floors, not performance claims. They exclude on-demand credits and the developer time required to specify, inspect, test, and merge the work. They also do not promise that the plan quota can complete that many tasks, because Devin does not publish the numeric allowance.

The useful formula is:

Plan price + on-demand credits + review cost, divided by accepted outcomes.

Track rejected work separately. A session that consumes quota and produces code you discard is a cost with no outcome. That is why upgrading for more raw activity can make unit economics worse.

Hidden Devin costs and billing rules

The main hidden cost is uncertainty: Devin exposes plan prices but withholds the numeric self-serve allowances. Several smaller rules then decide whether that uncertainty becomes a harmless cap or a variable bill.

On-demand usage can outgrow the subscription

On-demand credits are prepaid dollars for work beyond the included quota. They can be topped up manually or refilled automatically. Admins can set auto-reload thresholds and default per-session spending limits.

Purchased credits roll over month to month and never expire, which is better than a use-it-or-lose-it balance. Auto-reload can still turn an agent stuck on a poorly scoped task into repeated charges. Start with auto-reload off or tightly capped, then raise the ceiling only after accepted-output data supports it.

Review and Automations can use a separate pool

On Teams, Automations and Devin Review draw from shared on-demand credits instead of a full seat's included quota. If the shared balance reaches zero, Automations stop and Devin Review falls back to its smart diff viewer.

This is easy to miss in a seat forecast. Five full seats may cover interactive agent work while automated review creates a separate variable line. Set review triggers deliberately instead of running every workflow at the highest frequency by default.

Windows sessions carry a usage premium

Devin's usage docs say a Windows session consumes approximately 9% more usage than an equivalent Linux session. The operating system is therefore a cost choice. Use Windows when the code or tooling requires it, not as an unexamined default.

Sleeping sessions do not consume usage. Devin now documents automatic sleep after 30 minutes of inactivity by default. Clear prompts and bounded tasks still matter because active planning, context gathering, browser work, code execution, and virtual-machine time contribute to usage.

Monthly pricing is the public plan-card baseline

Devin's live public pricing page shows monthly prices and no annual discount. There is no advertised annual saving to calculate or annual commitment to justify from the current plan card.

That makes a monthly pilot the rational route. If a signed-in checkout later presents a longer term, compare its upfront total with twelve monthly payments before accepting it.

Refunds are narrow and renewal is automatic

Cognition's platform terms say payments are non-refundable except where the terms expressly provide otherwise or the law requires it. A fee dispute must be submitted no later than 10 business days after payment.

Subscriptions renew for another period of the same duration unless cancelled, and cancellation must arrive before the renewal date to avoid the next charge. Do not treat cancellation as a route to a prorated refund.

Self-serve credits and Enterprise ACUs are different budgets

By default, Devin-billed self-serve usage consumes quota and then dollar-denominated on-demand credits. Enterprise usage consumes ACUs at a contract rate. The old habit of multiplying every Devin task by a public ACU price no longer describes the self-serve bill.

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

For one developer, Devin Pro and Claude Pro cost $20 month to month; GitHub Copilot Pro costs $10. Cursor starts at $8 for Grok-only Plus, while Premium costs $30 and supports separately billed third-party models. At five full team seats, Devin's advertised total is $80 + 5 × $40 = $280 monthly. Cursor Teams starts at 5 × $30 = $150, Claude Team Standard costs 5 × $25 = $125, and GitHub Copilot Business costs 5 × $19 = $95. These are subscription entry budgets before extra usage.

Devin's five-seat premium is $130 over Cursor's starting team budget and $185 over Copilot per month, or $2,220 over Copilot across twelve monthly payments. That saving is meaningful only if Copilot's completion and agent workflow can replace the autonomous delegation you wanted from Devin. The archived illustration below preserves the earlier seat-price comparison: its Cursor price is no longer current, and its Devin column excludes the $80 team-plan fee. Use the current totals above to budget.

Cursor: lower entry, separate third-party usage

Cursor starts at $8 per month for Plus, with Grok models only; Premium costs $30 and adds access to separately billed third-party models. Teams starts at $30 per user per month, so five seats start at $150, $130 below Devin's advertised $280 total.

Premium includes cloud agents, automations, code hosting, and AI code reviews. Its included usage pool is shared across Cursor and Grok and refreshes weekly; third-party models have no included allowance and are billed at API list price. Teams adds central billing, administration, analytics, privacy settings, and SAML/OIDC SSO.

Choose Cursor when its editor and agent workflow fit the work. Choose Devin when its managed delegation workflow is worth the premium. Cursor's lower entry price needs a model check: Plus is Grok-only, and even Premium adds a separate bill for third-party models. Compare the models you will actually use before treating the subscription difference as a saving.

Claude Code: lower team price, terminal first

Claude Code is included in Claude Pro at $20 monthly or $200 paid upfront annually, displayed as $17 per month. Claude Team Standard costs $25 per seat monthly or $20 per seat per month when billed annually.

Five standard seats cost $125 on monthly billing or $100 per month equivalent on annual billing. That is $155 or $180 less than Devin's advertised budget with five full seats.

Claude Code is a terminal and IDE coding tool, and its subscription allowance is shared with the rest of Claude. Heavy chat, research, or file work can therefore reduce the capacity left for coding. API credits are a separate fallback billed at standard API rates, so an enabled API path can turn a flat subscription into a variable spend.

Choose Claude Code when terminal control and the wider Claude product matter. Choose Devin when the autonomous cloud-agent workflow, collaboration surface, and managed sessions are worth the higher team bill.

GitHub Copilot: the lowest five-seat team budget

GitHub Copilot costs $0 for Free, $10 per user per month for Pro, and $19 per user per month for Business. Five Business seats cost $95, less than half the advertised $280 Devin Teams budget.

Copilot Free includes 2,000 completions per month. Pro currently includes unlimited completions and next-edit suggestions plus $15 in total monthly AI credits. Business includes 1,900 AI credits per user, and paid usage beyond the pool costs $0.01 per credit, as confirmed in GitHub's billing docs.

Copilot wins when familiar editors, GitHub-centered controls, completion, and bounded agent work cover the job. Its lower seat price does not buy the same posture as Devin's autonomous task delegation. If the team still performs every long-running migration manually, the cheaper subscription did not replace the target outcome.

The price decision is clean. Pick Copilot for the lowest five-seat team entry among these plans, Claude Code for terminal-first work, Cursor for an editor-first environment with deliberate model spending, and Devin for managed autonomous delegation. The deeper workflow differences are covered in Codex vs Claude Code vs Cursor.

Which Devin plan should you choose?

Choose the lowest Devin tier that can complete accepted work without creating an expensive interruption. Price alone cannot select the plan because the numeric quotas are not public.

Choose Free when Devin is an evaluation tool, inline edits and Tab are the main use, or a quota reset can wait. Public PR review also has a free path.

Choose Pro for one developer using Devin regularly. The SWE-2 offer was scheduled to end on October 16, 2026, so check current in-product charges and measure ongoing spending separately from promotional usage. Pro remains the default paid choice because the jump from $20 to $200 is too large to make on intuition.

Choose Max when a full month of usage shows more than $180 in recurring Pro overage, or when the daily cap repeatedly blocks work worth more than the premium. A single sprint is not a recurring pattern.

Choose Teams with flex seats for occasional collaborators who can draw from shared credits. Watch their per-person consumption even though the balance is pooled.

Choose Teams with full seats for regular users who need their own Pro-equivalent quota or Devin Desktop. Budget the $80 team-plan fee plus $40 per full seat, and consider promoting a flex user above $40 of recurring monthly credits when the seat quota can cover the same work.

Choose Enterprise when SSO, dedicated deployment, organization-level ACU controls, custom terms, or a dedicated account team is a requirement. Those are governance needs, not reasons to guess at a public discount.

What changed in Devin pricing in 2026

The April 2026 announcement cut the minimum team entry budget from $500 a month to $80, a $420 monthly reduction at launch. In its April 14, 2026 announcement, Cognition retired the old Core and Team plans and introduced the Free, Pro, Max, Teams, and Enterprise ladder. That historical $80 minimum should not be substituted for today's advertised $80 team-plan fee plus full-seat charges.

That change made a controlled team pilot possible without a $500 commitment. It also ended the old no-minimum Core structure for light users. Someone who wants ongoing paid use now selects a subscription and then uses dollar-priced on-demand credits beyond its quota.

Self-serve billing also moved away from presenting ACUs as the buyer's day-to-day unit. Enterprise contracts still use ACUs, while self-serve customers see included quota and dollar-denominated credits. The consequence is easier invoice language but weaker public forecasting because the numeric allowances remain undisclosed.

Ask Devin Deep Mode, higher-quality DeepWiki modes, and Devin Review also moved toward usage-based billing in the announcement. Current public DeepWiki generation and public GitHub PR review retain free paths, but automated and higher-compute workflows belong in the variable-usage budget.

The new ladder is cheaper to enter and harder to forecast precisely. SWE-2's free Desktop and CLI offer was scheduled to end on October 16, 2026; the pricing page still shows that deadline without replacement terms. A promotion is not a steady-state usage budget.

The Monday move

Use this week to measure accepted work and the charges shown after the scheduled SWE-2 deadline. Start on Free unless a $20 Pro pilot is already worth the work you need done. Before running SWE-2 in Desktop or CLI, check the current model charge and usage controls in your account. The goal is one defensible line item next Monday, not a larger quota purchased on faith.

1. 1

   Start on Free

   Use the $0 plan on bounded tasks that resemble real work. Record each accepted result, rejected result, quota stop, and the time a developer spent reviewing it.

2. 2

   Buy one month of Pro

   Upgrade to $20 only when the Free limit blocks valuable work or the paid workflow is worth evaluating. Run representative tasks in Desktop or CLI after checking their current charges. Keep auto-reload off or capped, and record spending separately from any earlier promotional usage.

3. 3

   Price the accepted output

   At month end, add the subscription, on-demand credits, and review cost. Divide by accepted tasks. Ignore activity that did not survive review.

4. 4

   Promote measured usage

   Move a solo user to Max only above the $180 recurring threshold or a documented daily-cap cost. On Teams, consider a full seat above $40 of recurring credits if its quota covers the work, or when Desktop access becomes necessary. Include the $80 team-plan fee in the organization budget.

The result is a budget that follows work: Free while interruptions are harmless, Pro for a regular solo workflow, Max for proven heavy use, and Teams for collaboration with deliberate seat assignment.

Devin pricing FAQ

How much does Devin cost per month?

Devin Free costs $0, Pro costs $20 per month, Max costs $200 per month, and the current Teams card advertises $80 per month plus $40 per full developer seat, for up to 200 users. Enterprise is priced by quote. Five full Teams seats therefore budget to $280 before extra usage. The billing docs still show the older minimum model, so confirm checkout terms.

Is Devin AI free or paid?

Devin is both. Free includes a light agent quota, limited models, unlimited inline edits, unlimited Tab completions, and access to Devin Review and DeepWiki. Pro costs $20. Its free SWE-2 offer in Desktop and CLI was scheduled to end on October 16, 2026. The pricing page still lists that deadline as of October 17 and gives no replacement offer; do not assume continued free use.

What is a Devin ACU?

An Agent Compute Unit is the billing unit used in Devin Enterprise contracts at the rate in the customer's order form. By default, Devin-billed self-serve use draws on included quota first and dollar-denominated on-demand credits after it, so an old public ACU price should not be used to forecast Pro, Max, or Teams.

Is Devin better than Cursor?

Devin is the better fit when its managed delegation workflow is the priority; Cursor suits developers who prefer its editor and agent workflow. Devin Pro costs $20 monthly. Cursor Plus starts at $8 with Grok models only; Premium costs $30 with third-party model usage billed separately. With five full seats, Devin's advertised Teams budget is $280 versus Cursor Teams starting at $150.

Is Devin worth it?

Pro is worth $20 when one accepted task or one avoided delay returns more value than the monthly fee. Max is worth $200 only when its larger allowance displaces more than $180 of recurring Pro overage or equivalent daily-cap delay. Teams is worth its premium when collaboration and controls are required.

Does Devin offer a student discount?

Devin does not advertise a standing student discount on its live public pricing page as of October 17, 2026. Students should start on Free and pay only when the light agent quota blocks work they value.

What is Devin's refund policy?

Cognition's platform terms say payments are non-refundable except where expressly provided or required by law. Fee disputes must be raised within 10 business days of payment, and cancellation must arrive before renewal to avoid the next subscription charge.

Did Devin pricing change in 2026?

Yes. Cognition announced the lineup on April 14, 2026, retiring Core and the old Team plan and introducing an $80 minimum at that time. SWE-2 launched on September 10, 2026, and its free Desktop and CLI offer was scheduled to end on October 16, 2026. As verified October 17, the pricing page still displays that deadline without replacement terms. Teams still advertises $80 plus $40 per full seat; self-serve overage uses dollar-denominated usage.

Before adding another developer subscription, get the AI Business Workflow Audit Checklist.