Devin pricing explained
IC Guides · AI Agents · Coding agents
Devin is a coding agent from Cognition. Its self-serve plans have fixed monthly prices, and a daily and weekly usage quota sits inside each one. The bill changes after the quota runs out, because Devin then draws on prepaid on-demand credits. This guide sets out how Devin’s documentation describes each plan, the Teams seat arithmetic, and what stays unpublished.
The short answer
A solo developer pays $20 for Pro or $200 for Max. Devin’s Teams sources disagree as checked 18 October 2026, the pricing page now lists $80 a month with unlimited team members and no separate seat price, while the self-serve billing docs we read describe an $80 minimum including $40 full seat charges. Budget conservatively from the pricing page and confirm the actual checkout charge. Past the included quota, work is paid from on-demand credits. Cognition publishes no quota sizes and no rate for converting credits into work, so the only way to find your real cost is a trial on your tickets.
Plans and seats
| Plan | Price | Members | What the docs say |
|---|---|---|---|
| Free | $0 | 1 | Limited Devin usage, Devin Review and DeepWiki |
| Pro | $20 a month | 1 | Daily and weekly quota across Devin sessions, Devin CLI and Devin Desktop. Slack, Linear and MCP integrations. On-demand credits past the quota. |
| Max | $200 a month | 1 | A much larger weekly quota with no daily cap, shared across the same tools |
| Teams | Pricing page, $80 a month, no seat price shown; billing docs, $80 minimum including $40 full seats | Pricing page, unlimited. Docs, as read 18 October, up to 200 | Full seats at $40 each, free flex seats, shared on-demand credits |
| Enterprise | Custom | Larger organisations | Billed in Agent Compute Units at the rate in the order form |
Pro and Max cannot be shared. Anyone who wants several people on one subscription must use Teams. The pricing page now says unlimited team members, while the billing documentation, as we read it on 18 October, listed up to 200 team users. Their Teams billing descriptions still conflict; confirm the seat allocation and charge before committing.
Two conflicting descriptions of Teams billing
The following table follows the self-serve billing documentation, which describes an $80 monthly minimum including full-seat charges. The pricing page now presents $80 a month with no separate seat price; these are alternative readings, not a confirmed checkout quote. A full seat costs $40 and carries a Pro equivalent quota plus Devin Desktop access. A flex seat is free, draws only from the shared credit pool, and has no Desktop access. When you hold fewer than two full seats, Devin charges the rest of the $80 as prepaid credits that everyone can use.
| Full seats | Credits included in the minimum | Monthly total before top ups |
|---|---|---|
| 0 | $80 | $80 |
| 1 | $40 | $80 |
| 2 | $0 | $80 |
| 3 | $0 | $120 |
| 5 | $0 | $200 |
| 25 | $0 | $1,000 |
The billing-docs interpretation gives five full seats at $200 and 25 at $1,000 a month. An earlier reading that added a separate $80 fee would have given $280 and $1,080, but the current pricing page shows no seat price. These figures are our arithmetic from the billing docs only, excluding extra usage; ask Cognition which applies to your account.
What consumes usage
Usage comes from the number and complexity of actions, such as planning, gathering context, running code and browser steps, plus virtual machine time and network bandwidth, which Devin says is typically a small share.
Windows sessions use about 9% more than equivalent Linux sessions, while macOS sessions currently match Linux, which Devin calls promotional launch pricing.
Devin sleeps after 30 minutes of inactivity by default and uses nothing while asleep. It also uses nothing while waiting for you, waiting for a test run or cloning repositories, apart from a few units to keep the machine running.
Devin lists these drivers of consumption, namely task complexity, prompt specificity, codebase size, files touched, session length, conversation length and how often you message back and forth. Scoped tasks with a clear end goal and short sessions cost less.
On-demand credits
On-demand credits are prepaid, roll over month to month and do not expire. They can be topped up at any time or refilled by auto reload, and admins can set default session spending limits. On Teams they are shared with no per member balance.
Credits have the same dollar value as the older ACUs that legacy plans used. If credits run out, Automations stop and Devin Review falls back to a plain diff viewer, although reviewing a public pull request through devinreview.com uses no credits.
Pro, Max and Teams users who link an eligible ChatGPT Go, Plus or Pro subscription can have GPT model usage billed there instead of against their Devin quota.
What Cognition does not publish
Cognition does not publish the size of the daily or weekly quota on any plan, how many credits a typical task such as a bug fix or a pull request review consumes, or the dollar rate of an on-demand credit unit beyond the statement that it matches the old ACU value.
Session Insights shows the usage of each session, and the usage page under Settings shows your quota and credit balance. Those two screens are how you build your cost per task.
Where costs surprise you
A flex seat looks free, but all of its usage comes out of the shared credit pool, so a team of ten flex users can drain credits faster than a few full seat holders. Long conversations with many back and forth messages cost more than a single well scoped request. Enterprise customers instead consume ACUs against an order form, with limits that block new work when an organisation reaches its cap.
Questions to ask before you buy
- How large is the weekly quota on the plan we would pick, in sessions or hours of typical work?
- What does one of our usual tickets cost in credits? Run three on a trial and read Session Insights.
- Who needs a full seat, and who can live on flex seats and the shared pool?
- What spending limits and auto reload thresholds should we set on day one?
- Is there a member cap on our plan, given the pricing page says unlimited and the docs said 200, and does Enterprise change the unit and the controls?
IC Lab · AI Agent Pricing Tracker
Vendor prices change. The tracker records when.
Every entry links to the vendor’s pricing page and shows the date we checked it. Open the tracker, or get major pricing changes by email every Friday in TalkHead.
Related guides Devin vs Replit · Claude Code vs Devin · Claude Code usage limits · Replit billing · AI agent glossary · Which AI agent, and what it costs
Sources checked 12 October 2026; pricing page rechecked 19 October 2026 and self-serve billing terms 18 October 2026, with their Teams charge conflict disclosed above Devin pricing page · Devin docs, self-serve plans · Devin docs, usage · Devin docs, Enterprise billing. Prices change, so check the pages before you buy. See the AI Agent Pricing Tracker for the vendors we track.
Read beyond this guide
Why AI workers challenge the per-seat SaaS model Read the business analysis behind this implementation decision. Compare Claude Code, Devin, Replit and Copilot costs Compare billing units and vendor documentation before choosing.