Devin
CognitionDevin is a hosted autonomous software-engineering agent that takes a task, works in a cloud development environment and delivers a pull request. It is reached through a web interface, a CLI, a desktop app and an API, and sessions can run on customer-managed machines through Devin Outposts. Cognition builds and sells it.
Openness
1 high confidence- source
- closed
- license
- proprietary(SaaS-only)
- self-host
- no
Devin is proprietary software sold as a subscription service, and its source is not published. Customers can run session execution on their own machines, but the agent loop stays in Cognition's cloud, so the product itself cannot be self-hosted.
- https://cognition.com/legal/platform-terms-of-service recorded 2026-09-29
The platform terms bar customers from source-level access to the service: users may not "reverse engineer, disassemble, decompile, decode, adapt, or otherwise attempt to derive or gain access to any software component of the Services".
- https://docs.devin.ai/cloud/outposts/overview.md recorded 2026-09-29
The Outposts page says customer machines run only command execution, file edits and repository access: "Devin's agent loop (inference and planning) continues to run in Devin's cloud". No source distribution is described.
Adoption
3 low confidenceDevin sells paid Individual, Teams and Enterprise tiers, evidence of broad commercial adoption, but Cognition publishes no user or seat count. Adoption here is read from that reported commercial traction rather than a counted download or user figure.
- https://devin.ai/pricing recorded 2026-09-29
The pricing page sells paid Pro, Max, Teams and Enterprise plans, including a team plan of "Up to 200 users" and "Deploy in your virtual private cloud (VPC)"; searched for a total user or seat count: none published.
Capability
3 medium confidenceDevin 2.0 scores 45.8% on SWE-bench Verified as a single unassisted agent on a third-party leaderboard. That is mid-tier, well behind OpenHands, which leads the same leaderboard with an open scaffold.
- https://awesomeagents.ai/leaderboards/swe-bench-coding-agent-leaderboard/ recorded 2026-09-29
The leaderboard says "Devin 2.0 sits at 45.8% in standard unassisted mode" and lists OpenHands with CodeAct v3 at "68.4%" on SWE-bench Verified.
- https://cognition.ai/blog/swe-bench-technical-report recorded 2026-09-29
Cognition report: "Devin successfully resolved 79 of the 570 issues, giving a 13.86% success rate", the 2024 result on a 570-problem subset.
Verified 2026-09-29