This capstone project from Agent Engineering Bootcamp introduces an On-Call Support Agent to assist engineers with application maintenance and operational support. Built using Slack, Claude Sonnet, LangChain, OpenAI and Chroma, the agent uses operational runbooks and GitHub repositories to provide relevant, context-aware responses to help engineers quickly resolve incidents and better understand the technical details of an application.
Goals
-
Accelerate incident response by providing on-call engineers with targeted resolution steps.
-
Answer system architecture questions by analyzing GitHub repositories.
-
Help onboard new engineers with an assistant that answers questions about the operational and technical details of an application.
Takeaways
-
Retrieval-Augmented Generation (RAG) with operational runbooks gives the agent a solid foundation for relevant and accurate responses.
-
The GitHub Model Context Protocol (MCP) server allows the agent to search and analyze a codebase in real-time to answer system design questions.
-
Slack commands helped to create purpose-built workflows, with distinct commands for incident response, repository analysis and private messaging.