As observability systems grow more complex, the cognitive load on users increases quite fast. This talk presents an approach that could be game-changer in the future: using AI assistants as intelligent interfaces to your observability stack. By implementing and using MCP (Model Context Protocol) servers, we can transform how observability users interact with metrics, logs, and traces. You will see how teams can query their stack in plain English and use natural language to explore data, debug issues, and even work with configurations.
The session covers both theoretical foundations and practical implementation. It demonstrates how you can integrate AI assistants directly into your day-to-day workflows and provides a comprehensive walkthrough of:
- MCP architecture and how it enables LLMs (Large Language Models) to execute observability tasks
- Setting up and configuring MCP servers (demonstrated with VictoriaMetrics) and integration with popular AI assistants
- Current and planned features of VictoriaMetrics MCP Server
- Real-world use cases: data exploring, query explanation, working with alerting rules, cardinality analysis, intelligent debugging, obtaining context-rich answer for your questions, etc
- Various tips on how to make AI assistants work better with the observability stack
Whether you're an SRE looking to reduce toil, a platform engineer seeking to democratize monitoring access, or a leader evaluating AI's role in operations, this talk provides practical insights and tools for possible transformation of your observability practice.
This approach doesn't replace monitoring expertise at the moment — it amplifies it, making expert knowledge accessible to entire teams, giving you a powerful teammate in the form of AI assistant.