Snowflake Observe adds CLI and redesigned MCP for AI agents to access telemetry directly

Observability platforms have historically been built for human operators—dashboards, queries, and alerts designed for engineers to read and act on. Snowflake's announcement reflects a fundamental shift: AI agents are now the primary consumers of telemetry data. Agents can investigate errors before an engineer opens a ticket, correlate failures across services without human intervention, and automate routine debugging tasks. This changes what observability infrastructure must do.

The redesign of the MCP server is instructive. The old version included a custom LLM intermediary that received queries and returned answers, which was expensive and introduced latency. Snowflake removed that layer entirely, letting agents connect directly to Observe's APIs and context graph. This was shaped by feedback from engineering teams already building custom agents—several had built workarounds because programmatic access was limited. The new architecture gives agents visibility into Observe's data structure and permission to explore datasets, enabling more accurate and efficient queries using OPAL (Observe's query language).

The parallel rollout of a full-parity CLI means the same operations available to agents are also available to humans at the command line. Engineers can now compose, automate, and extend observability workflows outside the UI, and agents can work in the same interface as developers—from Claude Code to interactive terminal sessions. The prebuilt skills library (for incident investigation, failure tracing, change validation) lower the barrier to entry, so both humans and machines can start using the tools immediately after configuration.