Using ACP + Deep Agents to Demystify Modern Software Engineering | The JetBrains AI Blog

Jacob Lee, a founding software engineer at LangChain, shares his experience of developing a custom coding agent using Deep Agents and the Agent Client Protocol (ACP) to better align with his workflow preferences. While initially a small project during his paternity leave, the endeavor evolved into a significant enhancement of his daily software development processes, effectively replacing his previous tool, Claude Code. By integrating the agent with JetBrains IDEs, Lee achieved seamless code editing and agent interaction, leveraging features like human-in-the-loop support, observability through LangSmith tracing, and the ability to switch between different AI models such as OpenAI's GPT-5.4 and open-source alternatives like GLM-5. This project not only improved his productivity but also provided valuable insights into the inner workings of AI agents, demonstrating that developers can customize and control their workflows for better efficiency and understanding.