Building for Agent-First Observability with Observe by Snowflake

AI agents are investigating incidents, pulling production context on demand, and running without being paged. Most legacy observability platforms weren’t built to support this shift to AI-native engineering. They expose telemetry through dashboards and interfaces designed for humans to click through, not for agents to query programmatically.

With these new use cases being placed on existing observability platforms, observability must now work for both humans and agents.

The UI remains valuable for engineers who need to explore and investigate directly. Agents, however, need access to the same data through APIs, MCP servers, and CLIs. Observe has built exactly that: a redesigned MCP server and a CLI that gives agents direct access to the same data your engineers use.

This webinar covers the GA release of the Observe MCP server and CLI. We will show how the MCP server works and what changes when your agents have direct access to production telemetry.

You will learn:

• Why AI-native engineering requires observability platforms designed for both humans and agents
• How the redesigned Observe MCP provides agents with direct access to Observe's knowledge graph and APIs, removing the LLM intermediary that added latency and cost
• How to connect agents and developer workflows to live production telemetry using the redesigned MCP server