Observe Launches MCP Server and CLI for AI Agents to Query Telemetry Data

Snowflake has announced the general availability of a redesigned Observe MCP server and a new Observe CLI, both enabling AI agents and engineers to connect directly to the observability platform's telemetry data. The rollout is happening now across all clusters, with the eu-2 and ca-1 regions to follow shortly. These tools are designed for a world in which AI agents—including coding assistants, AI SREs, and custom alert-triage agents—are primary consumers of observability data alongside humans.

The previous Observe CLI provided only narrow, disconnected access to the platform, as most functionality was locked behind the UI. The new CLI offers full programmatic access to Observe's capabilities from the command line, with identical functionality exposed through the MCP server and the web interface. This enables workflows that can run autonomously in the background to handle routine tasks, as well as interactive investigations where an engineer guides the process. Both tools ship with a library of prebuilt skills—structured workflows for common observability tasks including incident investigation, failure tracing, change validation, n+1 issue detection, and outlier spotting—that work immediately after configuration.

The most significant architectural change is the removal of an LLM intermediary that was built into the original MCP server. Previously, agents would send queries to a single endpoint where Observe's own LLM would interpret the question and return an answer. This design meant agents lacked direct visibility into Observe's data structure and could not explore datasets themselves, creating latency and cost overhead for both Observe and its customers. The rebuilt server eliminates this layer: agents now connect directly to Observe's APIs, the same ones that power the Observe frontend. This gives agents direct access to Observe's context graph, enabling them to determine which datasets are most valuable to query and write efficient OPAL queries against any telemetry data. Agents can now access APM service maps, OpenTelemetry data collection setup, and active alert listings—all the capabilities available in the UI.

Engineering teams using the tools are building custom alert-triage agents that automatically query Observe when an alert fires and pull production telemetry for debugging context, as well as copilots that work alongside humans during incident investigations. To get started, engineers log into their Observe tenant, navigate to "Manage account" → "MCP server details", and follow the instructions for their agent of choice. Snowflake positions this shift as part of a broader recognition that observability is fundamentally a data problem, requiring that observability data be accessible not just through a UI but through the APIs, MCPs, and CLIs that modern AI agents and developer workflows rely on.