Last updated: 2026-08-25
Cognition's Devin is now a unified AI software engineering platform, joining the original cloud coding agent with Windsurf, rebranded Devin Desktop in June 2026, and Devin Local, a from-scratch rewrite of Cascade. It runs primarily on an in-house coding model built by Cognition, with optional access to frontier models from OpenAI, Anthropic, and Google on paid plans. A free plan and several paid tiers cover solo developers through enterprise deployments.
About Devin
Devin is an autonomous AI software engineer built by Cognition, a company founded in August 2023 by Scott Wu, Steven Hao, and Walden Yan. Launched in March 2024, Devin was marketed as the first AI agent that works as a software engineer rather than as a code-completion tool, resolving 13.86% of real GitHub issues end to end at launch, far above the previous best of 1.96%. It runs inside its own sandboxed virtual machine with a shell, code editor, and browser, so it never executes code directly on a user's computer. Devin operates as a compound system rather than a single model: a planner breaks down the task, a coder writes the implementation, and a critic reviews the result for bugs and security issues before it is shipped. The 2026 version runs on Claude Sonnet 4.5, which Cognition reports brought an 18% improvement in planning and a 12% gain in end-to-end evaluation scores, and added dynamic re-planning so Devin can change strategy mid-task if it hits a roadblock. It also accepts multi-modal input, including UI mockups, Figma files, and screen-recording video, to reproduce and fix visual bugs. Devin's standout use case is legacy code migration: it can ingest codebases written in COBOL, Fortran, or Objective-C and refactor them into Rust, Go, or Python while preserving business logic. It is also used for bounded engineering work such as dependency upgrades, lint fixes, test backfills, and bug fixes, where it opens a pull request, responds to review comments, and can record a QA walkthrough of its own testing. Nubank reported a roughly 12x improvement in engineering hours on migration work after adopting Devin. Pricing has three tiers. The Core plan is $20/month plus $2.25 per Agent Compute Unit (ACU), where one ACU is roughly 15 minutes of active autonomous work covering VM time, inference, and networking. The Team plan is $500/month and includes 250 ACUs at $2.00 each. Enterprise pricing is custom and adds VPC-isolated deployment, SSO, and SOC 2 Type II controls. Devin is delivered through Slack, Microsoft Teams, a web interface, and a VS Code extension. Cognition has grown quickly around Devin: it acquired the Windsurf AI coding IDE in July 2025, raised $400 million at a $10.2 billion valuation in September 2025, and was in talks for a $25 billion valuation by April 2026. Customers reported by 2026 include Dell Technologies and Cisco Systems, alongside SOC 2 Type II certification held since September 2024.
Screenshots
Pricing
Devin now runs a five-tier model instead of the original Core/Team/Enterprise ACU pricing: Free at $0/month (light agent-usage quota, unlimited inline edits and Tab completions, limited model access), Pro at $20/month (full model access including Cognition's own SWE-1.7 plus OpenAI, Anthropic, and Google frontier models, higher quotas, Devin Cloud agents), Max at $200/month (significantly higher quotas than Pro), Teams at $80/month base plus $40/month per full developer seat (unlimited members, centralized billing), and custom-priced Enterprise (VPC-isolated deployment, SAML/OIDC SSO, dedicated support). Usage beyond a plan's quota bills at API pricing. This replaced the earlier $20 Core-plus-$2.25-per-ACU and $500/month Team structure after Windsurf was folded into the Devin brand as Devin Desktop in June 2026.
Feature Comparison by Tier
| Feature | Core | Team | Enterprise |
|---|---|---|---|
| Base price (monthly) | $20 | $500 | Custom |
| Included Agent Compute Units | None (pay-per-use) | 250 ACUs | Custom |
| Per-ACU cost | $2.25 | $2.00 | Custom |
| Sandboxed VM execution | ✓ | ✓ | ✓ |
| MCP Marketplace integrations | 200+ | 200+ | 200+ + custom |
| Parallel agent sessions | ✓ | ✓ | ✓ |
| Multi-modal input (Figma, video, mockups) | ✓ | ✓ | ✓ |
| VPC-isolated deployment | — | — | ✓ |
| SSO / SAML | — | — | ✓ |
| SOC 2 Type II compliance | — | — | ✓ |
Key Features
- Autonomous end-to-end task delivery: Devin plans, writes, tests, debugs, and opens a pull request for scoped engineering tasks without step-by-step supervision, completing tasks like a pagination bug fix with a passing test in about 12 minutes.
- Sandboxed VM environment: Each session runs in an isolated virtual machine with its own shell, code editor, and browser, so Devin never executes code on the developer's local machine.
- Legacy code migration: Devin ingests large legacy codebases written in COBOL, Fortran, or Objective-C and refactors them into Rust, Go, or Python while preserving business logic, cutting migration engineering hours by roughly 12x at Nubank.
- Multi-modal bug triage: Devin accepts UI mockups, Figma files, and screen-recording video as task input, letting it reproduce and fix visual bugs that text descriptions alone can't capture.
- Self-QA with recorded walkthroughs: Before requesting review, Devin can run the app, click through the affected flow, and send back a recorded video of its own testing for a pull request.
- MCP marketplace: Devin connects to 200+ external tools including Linear, Datadog, Slack, Figma, Stripe, Zapier, and Airtable through the Model Context Protocol, plus custom STDIO, SSE, or HTTP servers.
- Parallel sessions: Since February 2026, Devin can run multiple agent sessions on different tasks at the same time with improved context retention on long-running work.
- Unified Devin Desktop, Cloud, and Local: Windsurf was rebranded Devin Desktop and Cascade was rewritten from scratch as Devin Local, so one Devin surface now spans an IDE, cloud task delegation, and a local agent instead of three separate products.
Pros
- Resolved 13.86% of real GitHub issues end-to-end at launch, about 7x the prior best of 1.96%, and Nubank reports a roughly 12x engineering-hours improvement on migration work using Devin.
- Entry pricing dropped 96% from $500/month to $20/month in 2026, lowering the barrier for individual developers to try autonomous task delegation.
- Sandboxed VM with its own shell, browser, and editor lets Devin run multi-hour tasks such as legacy migrations and environment setup without touching a developer's local machine.
- MCP marketplace gives access to 200+ tool integrations, including Linear, Slack, Datadog, and Figma, without custom integration code.
Cons
- An independent test found Devin completed only 3 of 20 assigned tasks, well short of the 'fully autonomous' framing used in marketing.
- Usage is billed in Agent Compute Units at $2.00-2.25 each (about 15 minutes of work per ACU), so real monthly spend commonly lands at $300-500 even on the $20 Core plan.
- Devin's only published SWE-bench Verified score (13.86%, from its 2024 launch) is far behind 2026 frontier models such as Claude Mythos 5 at 95.5%, and Cognition has not published updated figures for Devin 2.0.
- Refactoring can be superficial, moving code into new files without properly separating concerns, and Devin lacks the architectural judgment needed for major redesigns.
Product Information
- Cloud
- Yes
- Self-Hosted
- No
- On-Premise
- No
- Languages
- English
- Training
- Official documentation, Onboarding webinars, Community Discord, Video tutorials, API reference guides
Data Handling
- Training-data policy
- Customer code and data are not used to train Cognition's models. Enterprise deployments run within the customer's own VPC with data kept in the customer's controlled environment.
- Compliance
- SOC 2 Type II
Frequently Asked Questions
What is Devin and what does it do?
Devin is an autonomous AI software engineer built by Cognition, a company founded in August 2023 by Scott Wu, Steven Hao, and Walden Yan; launched in March 2024, it was the first AI agent marketed as a full software engineer rather than a coding assistant. It works inside a sandboxed virtual machine equipped with a private shell, editor, and browser of its own, so it never runs code directly on a user's computer. Devin can plan a task, write the code, run tests, debug failures, and open a pull request without step-by-step supervision; in 2026, Cognition added dynamic re-planning, so if it hits a roadblock it can change strategy mid-task instead of stalling. By 2026, Devin's customers included Dell Technologies and Cisco Systems, and Nubank reported a roughly 12x improvement in engineering hours on migration work.
How much does Devin cost in 2026?
Devin's 2026 pricing reset replaced the original Core/Team/Enterprise ACU structure with five simpler tiers, from a $0/month Free plan through a custom-priced Enterprise plan. Pro costs $20 a month for full frontier model access and Devin Cloud agents, while Max carries significantly higher quotas above Pro and Teams runs $80 a month plus $40 per full developer seat with unlimited members. Usage beyond a plan's included quota bills separately at API pricing, and Enterprise adds VPC-isolated deployment, SAML/OIDC SSO, and dedicated support.
What are the main features of Devin?
Devin's core feature is end-to-end autonomous task execution: it plans, writes, tests, debugs, and opens pull requests for scoped engineering work, running the whole time in a sandboxed VM with a shell, editor, and browser that isolates execution from the user's machine. A standout 2026 capability is legacy code migration: Devin ingests old COBOL, Fortran, or Objective-C codebases and rewrites them in Go, Python, or Rust, preserving the underlying business logic throughout. Devin also accepts multi-modal input, including UI mockups, Figma files, and screen-recording videos, to reproduce and fix visual bugs, and before asking for review it records its own walkthrough of the fixed flow as proof of testing. Since February 2026, Devin supports parallel sessions, letting it work on multiple tasks at once with improved context retention on long-running work.
Is Devin free to use?
Yes: since the 2026 Devin Desktop unification, Devin has a genuine Free plan at $0/month with a light quota for agent usage plus unlimited inline edits and Tab completions, though model access is limited compared to paid tiers. Pro at $20/month adds full frontier model access, higher quotas, and Devin Cloud agents, with extra usage beyond the included quota billed at API pricing. This replaces the earlier structure, where the cheapest plan, Core at $20/month, had no free tier and billed all agent work separately in Agent Compute Units.
What are the best alternatives to Devin?
Claude Code is the strongest alternative for teams that want deep repo-wide reasoning, multi-file refactoring, and terminal-based agentic coding at flat subscription pricing rather than usage-based billing. Cursor is the best choice for developers who want an interactive AI-native IDE for daily pair-programming rather than asynchronous task delegation. GitHub Copilot remains the safest rollout for large organizations already standardized on GitHub, with multiple distinct agent surfaces as of 2026. Choose Devin specifically when you want a task handed off completely and returned as a reviewable pull request, and choose the alternatives above when you want to stay in the loop during coding.
Who is Devin best for?
Devin is best for engineering managers and platform teams with a backlog of well-scoped, bounded tasks such as lint fixes, dependency upgrades, test backfills, and bug fixes with clear acceptance criteria; it is particularly strong for legacy code migration projects, where companies like Nubank have used it to cut migration engineering hours by roughly 12x. Enterprises that need SOC 2 Type II controls, VPC-isolated deployments, and SSO for autonomous coding agents are a good fit for the Enterprise plan. Devin is a weaker fit for solo developers on a tight budget who need every task handled inside the Free plan's light quota, since meaningful autonomous work quickly requires the Pro plan or extra API-priced usage. It also struggles with ambiguous product requirements and complex architectural decisions, where it can fall into unproductive debugging loops, so the right mental model is a capable junior engineer who needs a clear brief and PR review, not a senior architect working independently.
Does Devin have an API?
Yes, Devin exposes an API with keys prefixed apk_ that lets external systems delegate tasks to Devin and monitor execution, and this powers community projects such as the devin-mcp server; it also reaches the same third-party tools available through its built-in MCP Marketplace via the Model Context Protocol. If a needed integration isn't in the marketplace, users can add any custom MCP server over STDIO, SSE, or HTTP transports. By Q2 2026, the broader MCP ecosystem included community servers for GitHub, PostgreSQL, Docker, and Kubernetes that Devin can connect to. Task delegation itself is primarily handled through Slack or Microsoft Teams, where assigning Devin a task in a channel spins up a new session, which is what makes the API and MCP support useful as part of a larger automated pipeline rather than only a chat-based tool.
How does Devin compare to Claude Code in 2026?
Claude Code is generally rated the strongest overall AI coding tool in 2026 for repo-wide reasoning, multi-file refactoring, and terminal-based agentic work, and it runs on a flat subscription rather than usage-based billing. Devin, by contrast, is built for asynchronous delegation: you hand it a scoped task via Slack and it returns a pull request, working in its own sandboxed VM with a shell, editor, and browser. Devin now runs primarily on SWE-1.7, Cognition's own coding model, within a quota-based Free-through-Max pricing structure, a shift from the Claude Sonnet 4.5 foundation and ACU billing Devin 2.0 launched on. Choose Claude Code if you want an interactive agent for complex same-session engineering work, and choose Devin if you want tasks fully delegated and returned as reviewable PRs without your involvement during execution.
Top Alternatives
- Claude Code: Choose Claude Code for repo-wide reasoning and multi-file refactoring on flat subscription; choose Devin when you want tasks fully delegated as reviewable pull requests.
- Cursor: Pick Cursor for an interactive AI IDE you drive in real time; pick Devin when you want autonomous task delegation with Slack/Teams integration.
- GitHub Copilot: Choose Copilot for inline code completion in your IDE; choose Devin when you need a fully autonomous agent to plan, code, test, and open pull requests.
HokAI guides covering Devin
- Devin Split Into Three Products in June. One Stayed Vulnerable Until August.: Cognition renamed Windsurf to Devin Desktop on June 2, 2026. The rebranded product carried an unpatched GhostApproval flaw for 24 days, until an Aug. 1 fix.
- Windsurf Is Now Devin Desktop: What Actually Changed: Cognition renamed Windsurf to Devin Desktop on June 2, 2026, via a silent update. Pricing held steady, but the default screen and Cascade's future did not.
- Best AI Coding Assistants in 2026: Pick the Job, Not the Brand: Cursor, Claude Code, Devin Desktop, Tabnine and Trae compared by job, not brand: which coding tools to skip, verified against vendor pricing in August 2026.
- We Opened Every Source Link HokAI Cites. 531 of 535 Answered.: HokAI publishes the sources behind every guide. On 19 August 2026 we requested all 535 addresses they cite and 531 returned a live page. The method, repeatable.
- Claude Max Used to Die by Wednesday. Anthropic's Fix Expires August 19.: Anthropic doubled Claude Code's limits in May and boosted them 50% more since. That extra 50% is a promotion, extended four times, due to end August 19.