Tailkits product guide

Devin

Cognition's cloud coding agent that tests in its own VM and opens PRs, from $20

FreemiumAI Coding
Visit Devin
Website
devin.ai
Pricing
Freemium
Categories
AI Coding
Listed
Sep 2026
Updated
Oct 2026

Details about Devin

Short verdict: Devin is Cognition's coding agent: cloud agents that each work in their own virtual machine, test their changes in a browser and open pull requests, sold on one set of plans with Devin Desktop (the IDE formerly called Windsurf) and the Devin CLI. It is built for delegation: tag @Devin in Slack, Microsoft Teams, Linear or Jira, or let automations pick up alerts and failed CI runs. The catches are unpublished quota sizes, a free plan without the cloud agent, and training on your code unless you pay and opt out. It suits teams with a backlog of scoped tickets, not developers who want a fixed allowance or a no-training default.

How we reviewed it: We did not sign up for Devin, install Devin Desktop or the CLI, or give it any code. On 21 October 2026 we read the Devin pricing page, the billing, usage, security, models and Devin Desktop documentation, the release notes and changelogs, Cognition's blog and its Platform Terms of Service. Plans and limits change often; check the Devin pricing page before you pay.

What Devin costs and includes

ItemDevin on 21 October 2026
TypeCloud coding agent ("autonomous software engineer") with an AI IDE and a terminal agent, made by Cognition
Where it runsWeb app at app.devin.ai; Devin Desktop (formerly Windsurf) for macOS, Windows and Linux; Devin CLI for macOS, Linux, WSL and Windows; Slack, Microsoft Teams, Linear, Jira, GitHub, GitLab and Bitbucket; JetBrains, Zed and Xcode through the Agent Client Protocol
ModelsCognition's SWE-2 and older SWE models, models from Anthropic, OpenAI and Google, and open models such as DeepSeek, Kimi and GLM; Fusion pairs a frontier model with a cheaper "sidekick"
Free tier$0: a light agent quota, limited models (SWE-1.6, per the docs), unlimited Tab completions and inline edits in Devin Desktop; no Devin Cloud, API or Ask Devin
Paid plansPro $20 a month (cloud agents, all models, up to 10 concurrent sessions); Max $200 a month (larger weekly quota, no daily cap); Teams from $80 a month, with $40 full seats and free flex seats, up to 200 users; Enterprise custom, billed in Agent Compute Units
How usage is limitedDaily and weekly quotas whose sizes Cognition does not publish; past them, prepaid on-demand credits at API prices that never expire
Open sourceNo; the CLI installs with a script, Homebrew or a Windows installer, not npm
Data useCognition may train on customer data; paid plans can opt out, which also turns on zero data retention with model providers; the free plan cannot opt out
Latest notable releaseDevin Oncall generally available (15 October 2026): it watches Slack, Teams and PagerDuty, triages alerts and posts a likely root cause

Where Devin is different

Every cloud session gets its own computer

Each Devin Cloud session runs in its own virtual machine. Cognition says Devin starts Windows, Mac or Linux machines as needed, tests its work in its own browser and records a video of the result. Usage counts Devin's actions plus machine time; Windows sessions cost about 9% more than Linux ones. Sessions sleep after 30 idle minutes, use nothing while asleep and can be resumed for 14 days.

Work arrives from Slack, Teams and your tracker

Besides Slack and Teams threads and Linear or Jira tickets, automations can start Devin on a schedule or when a bug report, failed CI run or PagerDuty alert arrives. Devin can split a large task across "managed Devins" in separate machines. Devin Review checks pull requests for bugs and vulnerabilities, and it is free for public GitHub pull requests at devinreview.com.

Devin Desktop (formerly Windsurf) and the CLI on the same plan

Cognition agreed to buy Windsurf in July 2025 and renamed the IDE Devin Desktop on 2 June 2026; windsurf.com now redirects to Devin's site. It is the same VS Code-derived editor, with unlimited Tab completions on every plan and an Agent Command Center board for local and cloud sessions. Since 8 September 2026 its only built-in agent is Devin Local, which the CLI also uses. Cognition says Devin Local is up to 30% more token-efficient than Windsurf's old Cascade agent, and it adds subagents, sandboxing, plan mode and worktrees. On paid plans, Codex CLI, Claude Agent and OpenCode can also run inside Devin Desktop, and the CLI's /handoff command passes a task to a cloud Devin.

Limits and costs to watch

  • Unpublished quotas. Free gets a "light quota" and Max "significantly higher quotas", with no numbers, and each message costs more or less depending on model, task and reasoning. Overage is billed at API list prices (SWE-2: $3 per million input and $15 per million output tokens).

  • The free plan is mostly an editor. It has no cloud agents, and Cognition may train on its data with no opt-out. On Teams, only an admin can opt out.

  • Repricing in 2026. In April 2026 Cognition dropped the no-minimum Core plan and began charging for Ask Devin's Deep Mode, Devin Review and premium DeepWiki. The pricing page still says "Free use of SWE-2" on Pro; the models documentation says that ended on 15 October 2026.

  • Team billing. Two $40 full seats cover the $80 Teams minimum; otherwise the rest becomes prepaid credits, which flex seats share. When credits run out, automations stop. SSO is Enterprise-only.

  • Task size. Cognition's docs say Devin can most likely do what a person could do in about three hours, and warn that it can still introduce bugs or insecure code.

  • Windsurf leftovers. Cascade is gone, the Windsurf JetBrains plugin is in maintenance mode, and one-click App Deploys to Netlify do not work with Devin Local.

Pros and cons

Pros

  • Cloud sessions run in their own machines, test in a browser and return a pull request with a recording.

  • One $20 plan covers cloud agents, the Devin Desktop IDE and the CLI, with up to 10 parallel sessions.

  • Takes work from Slack, Teams, Linear, Jira and alerts, and runs scheduled automations.

  • A paid opt-out stops training and turns on zero data retention with model providers.

Cons

  • Quota sizes are not published, and overage is billed at API prices.

  • The free plan has no cloud agent and no training opt-out.

  • Frequent changes in 2026: new plans in April, the Windsurf rename in June, Cascade removed in September.

  • Closed source, and Cognition scopes it to tasks a person could do in about three hours.

How Devin compares

  • GitHub Copilot also turns an assigned issue into a pull request, from the $10-a-month Pro plan (1,500 AI credits, about $15 of usage), and works in more editors, including Visual Studio, Neovim and Eclipse. Choose it if your work lives on GitHub and you want a cloud agent for less.

  • OpenAI Codex is included in ChatGPT plans; from Plus at $20 a month you get cloud tasks that keep running while your computer sleeps, and its CLI is open source under Apache-2.0 and can run local models. Choose it if you already pay for ChatGPT.

  • Cursor is an editor-first alternative to Devin Desktop: its cloud agents also run in isolated VMs and return pull requests with screenshots or videos, but they need the $30 Premium plan, and Claude, GPT and Gemini usage is paid separately at API prices. Choose it if an editor is your main workspace.

See how Devin sits among all AI coding tools.

Who should use Devin

Devin fits engineering teams that work in Slack or Teams and a ticket tracker and have a steady queue of scoped work, such as bug fixes, dependency upgrades, migrations and incident triage, where a reviewed pull request is the deliverable. Former Windsurf users keep the same editor under a new name. Look elsewhere if you want a published monthly allowance, a free tier that includes the agent or a no-training default without paying.

FAQ

Is Devin free?

Partly. The Free plan costs $0 and includes Devin Desktop with unlimited Tab completions and a light agent quota, but not Devin Cloud, the autonomous cloud agent; that starts with Pro at $20 a month. Max costs $200 a month and Teams starts at $80 a month (prices as of 21 October 2026).

What happened to Windsurf?

Windsurf was renamed Devin Desktop on 2 June 2026, almost a year after Cognition agreed to buy it. It is the same IDE, now sold on Devin's plans, and windsurf.com redirects to Devin's site; Windsurf subscribers who paid $15 a month for Pro or $30 a seat for Teams keep those prices. Windsurf's Cascade agent was removed on 8 September 2026, and Devin Local, which the Devin CLI also uses, is now the built-in agent.

Does Devin train on my code?

It can by default. Cognition's Platform Terms say it may use customer data for model training; paid subscribers can opt out (on Teams, only an admin can), and opting out also turns on zero data retention with model providers. The free plan has no opt-out, and Enterprise data is not used for training without written consent.

How much does Devin cost for a team?

Teams costs at least $80 a month for up to 200 users as of 21 October 2026. Full seats cost $40 a month each and include a Pro-sized quota and Devin Desktop; flex seats are free and draw on the team's shared on-demand credits. Enterprise pricing is custom and billed in Agent Compute Units.