Connect AI Agents to Telemetry with Observe MCP & CLI
Snowflake’s Observe platform now offers a redesigned MCP server and a new CLI with full parity, enabling AI agents to directly query telemetry data for debugging and automation without human intervention.
Snowflake has announced the general availability of a redesigned Observe MCP server and a new CLI, both designed to provide programmatic access to telemetry data. The tools allow AI agents to interact with Observe’s platform directly, bypassing traditional UI workflows. Engineers can now automate tasks such as querying production telemetry during incidents or building custom alert-triage agents. The new CLI and MCP server offer the same functionality as the Observe UI, enabling agents to perform tasks like tracing failures or validating changes autonomously or interactively.
The previous Observe CLI covered only a limited set of platform features, restricting terminal-based workflows. The new CLI provides full parity with the MCP server, allowing users to compose, automate, and extend observability workflows from the command line. It supports both autonomous background tasks and interactive sessions, working seamlessly in environments like Claude Code. The tools include a library of pre-built skills for common observability tasks, such as incident investigation and detecting outliers, which require no additional setup after configuration.
The redesigned MCP server removes the need for an LLM intermediary, reducing latency and operational costs. Previously, agents had to route queries through a single endpoint where an Observe LLM interpreted requests, limiting direct data access. The new architecture allows agents to connect directly to Observe’s APIs, enabling them to query the platform’s context graph and write precise OPAL queries. This change eliminates the previous cost structure and provides agents with full visibility into Observe’s data structure and permissions.
To access the new Observe MCP server, users must log into their tenant, navigate to 'Manage account' → 'MCP server details,' and follow the setup instructions for their chosen agent. The platform’s shift reflects a broader trend where observability is increasingly consumed by AI agents rather than humans. Snowflake emphasizes that observability is fundamentally a data problem, and these updates aim to make telemetry data directly accessible to the AI-driven workflows teams already depend on.