Devin
Cognition · Autonomous Cloud AI Software Engineer
Devin is Cognition's autonomous AI software engineer: you assign it tickets from Slack, Microsoft Teams, Linear, Jira, the web app, or the API, and it works in its own cloud VM with a shell, editor, and browser to open pull requests. It is aimed at engineering teams delegating backlog work, reviews, migrations, and recurring chores.
Best for
Engineering teams with a steady backlog of well-scoped tickets who want to delegate them asynchronously from Slack, Teams, or their issue tracker and review the resulting PRs
Not ideal for
Developers who want interactive, in-editor pair programming, teams that require open-source or fully self-hosted agents, or anyone needing flat, predictable per-task costs
Who it's for
Engineering managers and software teams delegating tickets, reviews, migrations, and recurring maintenance to an autonomous cloud agent
Capabilities
- Each session runs in a cloud workspace with a shell, embedded IDE, and browser that you can watch and take over
- Start work by tagging @Devin in Slack or Microsoft Teams threads, assigning Linear tickets, or from Jira, the web app, and the API
- Devin Review organizes large PR diffs, flags likely bugs and security issues, and can auto-review PRs on open or new commits
- Code Scans (/scan) split a broad goal across parallel agents and return a prioritized findings report that Devin can turn into pull requests
- Scheduled sessions and automations for recurring work such as daily QA, release notes, or CI failure fixes
- Devin CLI hands local sessions off to Devin Cloud with /handoff and connects to the cloud VM with devin ssh
- Outposts run session command execution, file edits, and repository access on your own VMs, containers, or Kubernetes
- Microsoft 365 connectors (Outlook, Calendar, OneDrive and SharePoint, To Do, Teams, Entra ID) using delegated per-user permissions
Limitations
- Cloud agent access is listed from the Pro plan ($20/month) up; the Free plan offers only a light agent quota
- Usage is metered by actions, VM time, and bandwidth against quotas and then prepaid credits, so costs for long or complex sessions are hard to predict
- Cognition's security docs state customer data may be used for model training by default; paid plans can opt out, and Enterprise requires written consent
- Not open source, and the agent loop always runs in Cognition's cloud, even with Outposts
- The docs' rule of thumb is that Devin can most likely handle tasks a human could do in about three hours, so larger projects are best split into smaller tasks or parallel sessions
- Devin Review support varies by git host, with only partial Azure DevOps support
Use cases
- Tagging @Devin in a Slack bug thread and getting a pull request back with a fix
- Assigning a batch of Linear tickets to Devin so routine fixes arrive as PRs for review
- Running a Code Scan to find unused code or performance problems across a large repository
- Scheduling a daily session that fixes failing CI runs or drafts release notes
- Splitting a multi-repository migration across a team of parallel Devin sessions
- Handing a long-running CLI task off to Devin Cloud before closing your laptop
Devin pricing
Free
$0
- Light quota to code with agents
- Limited model availability
- Unlimited inline edits and Tab completions (Devin Desktop)
Pro
$20/month
Billed monthly
- Daily and weekly quota covering Devin sessions, Devin CLI, and Devin Desktop
- Access to cloud agents
- Full model availability
- Extra usage at API pricing
Max
$200/month
Billed monthly
- Significantly higher quotas than Pro
- Weekly allowance with no daily cap
Teams
$80/month + $40/month per full seat
Billed monthly
- Up to 200 users
- Full seats include a daily and weekly quota plus Devin Desktop access
- Free flex seats draw on shared team credits
- Centralized billing and admin dashboard
Enterprise
Custom
- SAML/OIDC SSO
- Custom deployment options, including a dedicated single-tenant environment
- Usage metered in ACUs against an order form
Free tier limits: The Free plan has a light agent quota and limited models; cloud agent access is listed as a Pro feature.
Note: Usage is metered by the actions Devin takes plus VM time and bandwidth. Self-serve plans consume their included quota first, then prepaid on-demand credits that roll over month to month; Devin does not consume usage while waiting for your reply or sleeping after inactivity.
Technical specs
Available models
Devin vs. competitors
Devin vs. windsurf
Devin Desktop (formerly Windsurf) is Cognition's IDE for working alongside an agent in your editor; Devin is the asynchronous cloud agent you delegate tickets to, and the same subscription covers both.
Devin vs. openhands
OpenHands is open source and self-hostable with any model; Devin is a proprietary managed service with deeper built-in Slack, Teams, Linear, and PR review workflows.
Devin vs. github-copilot
GitHub Copilot is built around GitHub's editors, repositories, and pull requests; Devin is a standalone cloud agent with its own VM per session that also works with GitLab, Azure DevOps, and Bitbucket Data Center, and adds Code Scans and Outposts.
Devin vs. claude-code
Claude Code is an agent you usually run and steer in your own terminal; Devin is designed for handing off tasks to run unattended in the cloud and report back with PRs.
Frequently asked questions
What does Devin do?
Devin takes software engineering tasks — bug fixes, tickets, reviews, migrations, recurring chores — and works on them autonomously in its own cloud workspace with a shell, editor, and browser, returning pull requests for your team to review.
Is Devin the same as Devin Desktop or Windsurf?
No. Devin Desktop is Cognition's AI code editor, formerly called Windsurf. Devin is the cloud agent you delegate tasks to from Slack, Teams, issue trackers, the web app, or the API. One self-serve subscription covers Devin sessions, Devin CLI, and Devin Desktop.
How much does Devin cost?
Self-serve plans are Free, Pro at $20/month, Max at $200/month, and Teams at $80/month plus $40 per full developer seat; Enterprise is custom. Cloud agent access is listed from Pro up, and usage beyond the included daily and weekly quotas draws on prepaid on-demand credits.
Can Devin run on my own infrastructure?
Partly. Devin Outposts, available on Pro, Max, and Teams accounts, run command execution, file edits, and repository access on machines you control, but Devin's inference and planning stay in Cognition's cloud. Enterprise also offers a dedicated deployment that Cognition hosts in a customer-isolated, single-tenant VPC connected over AWS PrivateLink or an IPSec tunnel; a fully self-hosted option is not documented.
Does Devin have an API?
Yes. The Devin API (v3 is current) lets you create sessions, send messages, manage knowledge and playbooks, start code scans, trigger Devin Review on PRs, and set up automations and scheduled sessions, authenticated with personal access tokens or service-user API keys.
Integrations & fit
Alternatives to consider
OpenHands
Developers and platform teams who want an open-source, bring-your-own-model coding agent they can run and automate on their own machines or VMs
FreemiumGitHub (Microsoft)
Developers on GitHub who want AI completion, chat, and agents embedded in their existing IDE and GitHub workflow
FreemiumAnysphere
VS Code developers who want multi-file AI editing, autonomous cloud agents, and model flexibility — not a plugin layered on top of an existing editor
FreemiumAnthropic
Developers and teams who want deep codebase reasoning, multi-file changes, and persistent project instructions in a terminal-friendly or IDE-native workflow
PaidAbout Devin
Devin is the cloud agent in Cognition's lineup, distinct from Devin Desktop (the IDE formerly called Windsurf, listed separately in this directory). Instead of working beside you in an editor, each Devin session gets its own cloud machine with a shell, an IDE, and a browser; you hand it a task and can watch the work log or take over at any point. Work arrives from where teams already talk: tag @Devin in Slack or Microsoft Teams, assign or label a Linear ticket, pick up Jira issues, start sessions from the web app or the v3 API, or hand off a local Devin CLI session to the cloud with /handoff. Around the core agent, Cognition has added Devin Review for organizing and auto-reviewing pull requests, Code Scans that fan a broad goal out to parallel agents and return a prioritized report whose findings can be turned into PRs, scheduled sessions and automations, and connectors for Microsoft 365, Datadog, Sentry, and data warehouses. The docs frame its sweet spot as tasks a human engineer could finish in about three hours, with a team of Devins for larger migrations. One self-serve subscription covers Devin sessions, Devin CLI, and Devin Desktop; cloud agent access is listed from the Pro plan up, and usage is metered against daily and weekly quotas before drawing on prepaid credits, which makes costs depend on how much work each session does. Outposts can run command execution on machines you control, but the agent loop always stays in Cognition's cloud, and there is no open-source or fully self-hosted option.
Updates from Devin
Devin can be @-mentioned in Teams chats and channels and can listen to a channel to start work on its own, while a Microsoft 365 integration adds Outlook, Calendar, OneDrive and SharePoint, To Do, Teams, and Entra ID connectors with delegated per-user permissions.
Cognition added its next-generation SWE-2 model to Devin's agent selector as a research preview, and the Devin CLI can now create, steer, resume, and watch Devin Cloud sessions, including handing local work off with /handoff.
Code Scans, started from the web app with /scan, distribute an investigation of a broad engineering goal across parallel agents and synthesize the results into one prioritized report, whose findings Devin can then turn into pull requests.
Are you the founder? Claim this listing →