Observe by Snowflake has released a redesigned MCP server and new CLI that allow AI agents to query observability data directly from the command line and from agent environments.
The new tools expose the full Observe platform with the same operations available in the UI, enabling engineers to build custom alert-triage agents and incident investigation copilots that connect to production telemetry without the latency and cost of the previous architecture's built-in LLM intermediary.
What happened
Snowflake announced general availability of a redesigned Observe MCP (model context protocol) server and new Observe CLI tool that give AI agents direct programmatic access to the full Observe observability platform. Both tools expose the same operations and are now rolling out across all clusters, with eu-2 and ca-1 to follow shortly.
Why it matters
As AI agents increasingly read telemetry data (coding assistants investigating errors, AI SREs correlating failures across services), observability platforms must support agent workflows alongside human-readable interfaces. The new tools let engineers build custom alert-triage agents and incident investigation copilots that query production data directly, reducing latency and eliminating the cost of an LLM intermediary that the old architecture required.
What to watch
The new architecture removes a custom LLM layer from the old MCP server and connects agents directly to Observe's APIs. Both tools ship with prebuilt skills for common observability tasks (incident investigation, failure tracing, change validation, n+1 issue detection, outlier detection) that work immediately after configuration.
Observe by Snowflake has announced the general availability of a redesigned MCP server and a new CLI tool that provide AI agents with direct programmatic access to the full Observe observability platform. General availability is rolling out now across all clusters, with eu-2 and ca-1 regions to follow shortly.
The new tools address a fundamental shift in how observability data is consumed. As Observe notes in its announcement, agents now have a sophisticated understanding of operational data—coding assistants can investigate errors before an engineer opens a ticket, and AI SREs can correlate failures across services without being paged. This means observability platforms must support both human and machine readers of telemetry data. The previous Observe CLI covered only a narrow slice of the platform, as most functionality was inaccessible from the terminal. The new, agent-compatible CLI provides programmatic access to Observe's full capabilities from the command line, with the same surface available through the MCP server and in the UI. It works from agent environments like Claude Code as well as interactive terminal sessions.
The most significant change is architectural and economic. The previous MCP server contained a custom LLM harness that received user queries, performed reasoning, and returned answers. This design prevented agents from exploring datasets directly and added latency and cost to every interaction. The new MCP server removes the LLM intermediary entirely. Agents now connect directly to Observe's APIs, gaining access to the same context graphs and capabilities that power Observe's frontend. This design gives agents direct visibility into Observe's data structure and the ability to write efficient OPAL (Observe's query language) to run precise observability queries against any telemetry data. Through the MCP server, agents can access APM service maps, OpenTelemetry data collection setup, and active alert listing—all without the overhead of the previous architecture.
Both the CLI and MCP server ship with a library of prebuilt skills—structured workflows based on how Observe engineers solve common observability tasks. These skills enable agents and engineers to investigate incidents, trace failures, validate changes, find n+1 issues, and detect outliers, all without additional setup after configuration. Engineers are already using these tools to build 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. The CLI is designed to support both autonomous workflows that handle routine tasks in the background and interactive workflows requiring an engineer's direct guidance during investigations.
Observability platforms have traditionally been built for human operators—dashboards, UI visualizations, and manual query interfaces. As AI agents become integrated into incident response and debugging workflows, the data models and access patterns that work for humans no longer suffice. Coding assistants that investigate errors before an engineer opens a ticket, or AI SREs that correlate failures across services, require programmatic APIs and structured data access that the UI-first approach cannot provide efficiently.
Snowflake's redesign reflects direct feedback from engineering teams already deploying agents against Observe. Several teams had built their own workarounds because the previous MCP server lacked supported programmatic access. The original architecture's key bottleneck was economic and technical: a custom LLM intermediary that interpreted natural-language queries meant agents could not explore datasets directly, incurred latency on every call, and forced Observe to absorb the cost of running an LLM harness. By removing the LLM layer entirely and exposing direct API access, the new MCP server and CLI lower operational cost, reduce latency, and grant agents the fine-grained visibility they need to write efficient queries. This shift reflects the broader industry recognition that agents and humans will read the same observability data through different interfaces—CLI, MCP, and UI—and that platforms must optimize for both.
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · takes 30 seconds · unsubscribe anytime
Ask AI anything about this article. Q&As are published on this page for other readers too.
A researcher released WMTrace, an open-source forensic workbench for detecting and analyzing text watermarks e…
Dynatrace has signed a definitive agreement to acquire Arize AI in a cash and stock transaction valued at $915…
Deltix, an AI-driven testing tool, has entered open beta
Z.ai announced GLM-5.3, a model with ~750B parameters that has surpassed Moonshot AI's Kimi K3 and matched or…
Anthropic will offer a watermark detection API allowing third-party developers to identify text created by Cla…
A developer released Dictata v0.1.0, a Windows application that transcribes speech locally using Whisper (an A…
The AI news that matters, in one minute each morning.
Sign up free