Agenten-Kompass

AGENT PROFILES

Devin

Cognition · Coding agents

Software agent for scoped development tasks that writes, runs, and tests code. Cloud sessions can be assigned through the web, integrations, or the local CLI.

Develop software

Devin · Cloud and local coding agent

Product and access

Devin is Cognition's software agent for cloud sessions. Devin CLI works locally and can hand longer tasks to cloud Devin. Devin Desktop is the associated desktop workspace; Cascade remains a separate local agent there.

01

Core features

6 features from provider sources · sources reviewed on September 30, 2026. Availability and limitations are listed directly below.

  • Write, run and test code

    Devin handles development tasks in a configured environment and can make and verify changes independently.

    Source 1

  • Observe and take over the cloud workspace

    The IDE, shell, and browser show the work of a cloud session; you can take over in the embedded workspace.

    Source 1

  • Local work through the CLI

    Devin CLI works directly with local files and tools in the terminal.

    Source 2

  • Delegate tasks to cloud Devin

    The local CLI can hand longer-running tasks over to a cloud session.

    Source 1

  • Work on tickets and pull requests

    Documented use cases include bug reports, features, tests, reviews, and prepared pull requests.

    Source 1

  • Handle multiple tasks in parallel

    Multiple clearly scoped backlog tasks can run in parallel as separate sessions.

    Source 1
02

Requirements and limits

‘Still open’ means the reviewed sources do not answer this point clearly.

  • Requirement

    Prepare the repository and environment

    Devin needs access to the repository and a working development environment with the required dependencies and access permissions.

    Source 1

  • Requirement

    Define a clear task scope

    The provider recommends clear instructions, explicit acceptance criteria, and easily verifiable results.

    Source 1

  • Limitation

    Break down complex tasks

    The documentation recommends dividing more difficult projects into manageable steps and providing relevant context.

    Source 1
03

Practical tasks

Editorial examples based on documented features. No product tests performed.

01

Fix a bug from a ticket

Your input
A reproducible ticket, repository access, and specific acceptance criteria.
Possible result
A tested change with an explained approach and prepared pull request.
What to check
Review diff, tests, CI result, and actual user flow.
02

Hand off a local prototype to the cloud

Your input
A local repository, current working state, and a clearly limited follow-up task.
Possible result
A cloud session that continues the longer task.
What to check
Check the transferred state, permissions, session log, and result.
EDITORIAL EXAMPLE

Fix a reproducible bug from a ticket, run tests, and prepare a pull request for review.

Your possible prompt

Read the project rules. Investigate [error], propose a small change, and run the relevant checks. Show me the diff and any remaining uncertainties.

Getting-started idea based on documented capabilities. Not an executed product test. Adapt the task, data, and permissions.

04

What you should know first

Current Cognition documentation for cloud Devin and Devin CLI. Specific integrations and access depend on team and plan configuration.

Requirements & limitations

Keep tasks narrowly scoped, provide clear acceptance criteria, and review code, tests, and the pull request yourself. Access depends on the plan and team configuration.

Setup

Set up account, repository access, development environment, and clear acceptance criteria.

Pricing model

Check plan / usage. Check current terms and included quotas with the provider.

The classification indicates the scope of effort. It is not a price or performance commitment.

05

Sources from the provider

Editorial team: Michael Meierhoff · sources reviewed on . Provider information is not independent evidence of performance.

Read profile as Markdown

How we assess the sources