Regarding a specific example of how I use Windsurf, I engaged in a contextual conversation with the agent for a cross-cloud integration between AWS and Azure, leveraging a hybrid approach between cloud and Grok models to find the most cost-effective logic for the script. I provided the agent with the specific target environment, a virtual machine, and defined the deployment requirements. Windsurf helped me generate and refine the PowerShell script needed to bridge the two platforms, while ensuring the automation was optimized for performance and budget.

I use Windsurf to implement a lot of pipelines on GitHub Actions. I was working with the migration from Jenkins on a legacy application, and Windsurf helped me significantly with this migration, debugging all the pipelines and aiding me substantially with CI/CD.

All of the team has started working with Windsurf. We use rules for all the team so we have a standard procedure for deployment. This has helped with the agile development of the team, and with another application such as Linear, working with issues and incidents.

We were deploying one time a week before we started working with Windsurf, and with Windsurf, we are succeeding in a maximum of five releases in a week.

Windsurf is built on VS Code, which is the industry standard. This is a huge advantage for me because it allows for a seamless transition. I can keep using all my essential DevOps extensions for Azure, AWS, and Kubernetes without any compatibility issues. The user interface is familiar and high performance, meaning there is no learning curve. It feels like my professional environment, but with superpowers, thanks to the integrated AI Flow.

A standout moment was during a migration from Jenkins to GitHub Actions, converting complex Jenkins files into GitHub Actions workflows, which is typically a tedious process. Using Windsurf AI Flow, I was able to point the agent to my existing Jenkins pipelines. The AI did not just translate the code; it mapped the logic across the entire codebase, identifying environment variables, secret management requirements, and build dependencies. It autonomously generated the new GitHub workflow files and even suggested fixes for syntax mismatches that would have taken me hours to debug manually.

Windsurf is supporting MCP, which is an important standard for the integration with other tools. Most of the context tokens are consumed by documentation. We need a good structure for the agent to be optimized with this. Most of the time, our developers are wasting money because the agent does not save the context properly.

The tokens were fully consumed in the middle of the month. With the on-demand plan, we do not get extra tokens, so we have to downgrade our use of the agent models to a free one. When that happened, it increased the time of development because all the developers had to prompt the agents in a better way. However, that is more related to the decision on the plans. In five or six days, I consume all the tokens.