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

The shift reflects a broader maturity in AI agent architectures and the observability market. Rather than having agents interact with telemetry through a conversational LLM layer, Snowflake has recognized that direct API access is both cheaper and more reliable—agents no longer incur the latency and cost of an intermediary interpreting queries, and they gain full programmatic visibility into Observe's data structure. This design change was shaped by customer feedback: engineering teams had already built workarounds to access observability data programmatically, indicating market demand that the old UI-and-LLM model did not satisfy.

The introduction of prebuilt skills—structured workflows for common observability tasks like incident investigation and outlier detection—lowers the barrier for teams adopting agent-assisted incident response. By shipping these workflows alongside the API access, Observe is enabling both autonomous and human-guided workflows: agents can triage alerts and pull debugging context without human intervention, while engineers retain the ability to steer investigations interactively. The full parity between the CLI and MCP server means the same capabilities available to agents in Claude Code or other agent environments are also accessible to engineers in terminal sessions, treating observability as a data problem that multiple interfaces should serve equally.