Introducing Devin
Devin is your collaborative AI teammate, built to help ambitious engineering teams achieve more.
Devin is an autonomous AI software engineer that can write, run and test code, helping software engineers work on personal tasks or their team projects.
Devin can review PRs, support code migrations, create Dagster jobs, respond to on call issues, build web applications, and even personal assistant tasks like ordering your lunch on DoorDash so you can stay in the zone on your codebase.
Here are just some of the use cases the Cognition team and Devin’s early users have used Devin for:
- Code migrations and refactors
- Bug and issue triage
- Unit & integration tests
- Data engineering, analysis and visualizations
- Scraping and repetitive web research
Tag Devin on a Slack thread about a bug you’re discussing with coworkers, delegate a task in your IDE to save you from context switching, or hand Devin your todo list at the start of your day and return to draft PRs waiting for review. Devin is most effective when it’s part of your team.
Getting Access
If your company is already working with Cognition, you can request permissions with your Administrator or Cognition directly and access Devin via the web application preview.devin.ai.
If you do not have access and your name is already on our waitlist, we do small waitlist releases every week and invite users to join our Technical Preview along with our Discord and Slack communities.
General Product Features
The Devin Interface
Devin is designed to be a conversational user interface, and allows you to follow Devin’s development and planning process in the Workspace section.
In Devin’s Workspace, you’ll find:
Shell
a text-based interface where Devin can input commands that interact directly with a computer’s operating system.
Browser
watch Devin browse through documentation, test web applications it builds, download/upload information, and jump in to help Devin navigate through browsing tasks.
Editor
this is where Devin will write code. You can supervise, review and edit (see our VSCode Guide) the code Devin is writing.
Planner
this section provides you a running “to do” list that Devin laid out to complete the task. This gives you visibility into how Devin plans and thinks through solving a problem.
Demo
Limitations
Devin is an early product and has lots to learn. The Cognition team is working hard to improve Devin’s breadth of capabilities.
As you use Devin, keep in mind tasks that will be harder for it to execute successfully:
- Large-scale challenges: Devin generally performs better on smaller-scoped tasks and single file rather than multi-file edits. For more complex edits, we recommend breaking up the project into smaller, isolated tasks that you can run in separate sessions that build on each other.
- Consistency and reliability: Devin will sometimes get off-track. Sometimes you’ll be able to nudge Devin back on track, other times you may want to restart the session.
- UI-related aesthetics: Devin can build functional frontends, but will need your help with aesthetics. Devin doesn’t have great eyesight.
- Mobile development: Devin can help with mobile applications, but doesn’t have a phone to test with.
- Security: This is our top priority at Cognition, but we ask all users to exercise caution when sharing credentials with Devin or granting access to specific platforms. Any time you are sharing credentials with Devin, make sure you use the Secrets Manager or via the secure credential-sharing channel provided by Devin in the chat.
Feedback
We’re learning and our customers’ input is crucial! If you are part of the Technical Preview, please reach out via our Slack Community. For all other customers and partners you can share your feedback to support@cognition.ai or directly in the web app.
We log all feedback provided by customers and use it to make quick improvements to Devin and inform our product priorities and roadmap.
About Cognition
We are an applied AI lab building end-to-end software agents.
We’re building collaborative AI teammates that help ambitious engineering teams achieve more.