Grafana today announced the public preview of gcx, a new CLI tool that brings full observability—monitoring, alerting, and incident response—directly into the terminal and the agentic coding environments engineers increasingly rely on.
Engineers using AI agents like Cursor and Claude Code can now spot and resolve latency spikes, SLO breaches, or failures in minutes without leaving their command line, eliminating the context-switching that slows down modern development workflows.
"The way engineers work has fundamentally shifted toward AI-assisted coding in the terminal, but production visibility remained trapped in separate tools," said Tom Wilkie, CTO of Grafana Labs. "With gcx, agents and engineers alike can see what's actually happening in production and act on it—all from the same environment where they write code."
Background
AI coding agents have become highly effective at generating and modifying source code, but they operate with zero visibility into production systems. They can see code on a developer's machine but not whether the service is actually hitting its SLOs or experiencing a checkout latency spike.
This creates a new "visibility gap": agents pattern-match on source files and make decisions based on assumptions rather than real-time system state. Engineers then must jump into a browser-based dashboard to investigate—breaking their flow and slowing incident response.
What This Means
gcx closes this gap by giving agents direct read/write access to Grafana Cloud from the terminal. Engineers can instrument services with OpenTelemetry, define alerts and SLOs, push dashboards as code, and even stand up synthetic probes—all via simple CLI commands or natural-language instructions to their agent.
The result: what previously took a multi-day ticket now happens in a single session with an agent. "You point your agent at a service and ask it to bring observability up to standard," explains Wilkie. "gcx exposes the primitives it needs across the full lifecycle—from instrumentation to alerting to frontend monitoring—all in the terminal."
Early adopters report cutting incident response time from hours to minutes. The tool also supports everything as code: dashboards, alerts, SLOs, and checks can be pulled as files, edited locally with an agent, and pushed back—with a deep link into Grafana Cloud for human review.
Quick Start with gcx
gcx is available now in public preview. Developers can install it via Homebrew or download directly from the Grafana documentation.
- Install:
brew install grafana-cloud-cli - Authenticate:
gcx login(links to your Grafana Cloud account) - Start monitoring:
gcx service add --name my-service– instrumentation, alerts, and SLOs generated in minutes.
For agent workflows, simply instruct your AI coding assistant to use gcx commands or ask it to check production health before suggesting code changes.
Availability and Next Steps
The public preview is free for all Grafana Cloud users. Grafana Labs plans to add deeper Kubernetes monitoring, automatic remediation workflows, and expanded agent integrations based on early feedback.
"This is just the beginning," Wilkie said. "We believe every agent should have production context—gcx makes that possible today."