Integration of AI agents in software engineering

The software development and DevOps sectors are experiencing a significant shift toward agentic AI workflows. Major technology firms are expanding infrastructure to support autonomous agents; Atlassian reported that 77 percent of its internal pull requests were fully generated by AI agents, leading to the introduction of the Agentic Multiplayer Protocol (AMP) to manage identity and governance in multi-agent environments. DigitalOcean has also introduced new services to help teams monitor, detect errors, and optimize these autonomous systems.

As these agents move from providing suggestions to actively investigating alerts and drafting infrastructure changes, the industry is focusing on security and compliance. Experts are advocating for layered defenses, such as OS-level sandboxing and the Model Context Protocol (MCP), to mitigate risks like prompt injection and agent drift. Additionally, the availability of models like Anthropic’s Claude on Amazon Bedrock within AWS GovCloud allows these agentic coding workflows to be deployed in highly regulated defense and federal sectors.