Snowflake releases AI-agent tools for observability data

The shift toward AI agents as primary consumers of observability data represents a fundamental change in how operational telemetry is accessed and used. Snowflake's redesign reflects feedback from engineering teams already building custom agents for incident response—teams that had previously worked around the lack of programmatic access by building their own solutions. By removing the LLM intermediary from the MCP server architecture, Snowflake addresses both the cost and latency challenges that made the old approach expensive for both the company and its customers.

The full parity between the CLI and MCP server is significant because it means every operation agents can perform is also available to engineers in the command line, creating a unified interface across interactive terminal use, agent-powered automation, and the existing UI. The prebuilt skills—structured workflows for common tasks like tracing failures and detecting outliers—lower the barrier to adoption by eliminating the need for additional setup after configuration. This allows both humans and machines to begin using the tools immediately, addressing a core requirement for observability platforms that must now support dual workflows: some running autonomously in the background and others requiring direct human guidance.