Grafana releases gcx CLI and MCP server to let AI agents query live observability data during development

Grafana Labs has released two tools that integrate observability data into AI-driven development workflows. The gcx CLI and MCP server allow AI coding agents to pull live metrics, logs, traces, SLOs, and synthetic monitoring results from Grafana Cloud or self-hosted instances. This addresses a gap where agents generate code without understanding its real-world impact, relying instead on static training data or superficial code reviews. The tools provide an evidence-based check by grounding agent decisions in observed system behavior.

The MCP server offers a more opinionated approach, providing predefined tools for common use cases, while gcx is a flexible CLI that lets agents build custom workflows. Both tools work with Grafana Cloud or self-hosted OSS and Enterprise instances. Grafana Labs also provides installable skill bundles for gcx and separate plugins for the MCP server, enabling agents to interact with observability data without manual configuration. This reduces the overhead of integrating telemetry into agentic workflows.

In practice, agents can use these tools to validate implementation decisions against real-world data. For example, an agent adding a new payment provider can check current latency metrics to set realistic test conditions, rather than relying on assumptions. Agents can also read and modify existing dashboards, trace queries back to source code, and push updated definitions into Grafana or source control. This creates a feedback loop where telemetry informs development, rather than being a separate post-deployment concern.

For local development, agents can stand up an OpenTelemetry Collector or use Grafana’s otel-lgtm Docker image to export build telemetry to a local Grafana stack. The gcx CLI can pull production dashboard definitions into this local instance, while agents can generate realistic k6 load-test scripts from observed traffic. This reduces the manual effort of setting up test environments, which previously took about a day. Grafana Labs also highlights an internal use case where agents profile development environments to identify optimization candidates and measure improvements.

The tools also address front-end regressions, which backend telemetry often misses. Grafana Labs has introduced an experimental Agentic Testing feature that uses natural-language instructions to check UI flows in live web applications. This expands the scope of observability-driven development beyond backend systems, making it possible to catch issues that would otherwise go unnoticed until deployment.