Kiro IDE
Give Kiro's agent read-only access to your live production telemetry. Logs, traces, metrics and Kubernetes state answer questions inside the editor, on the same spec you are working on.
Overview
Kiro is an agentic IDE built around specs: you describe the change you want, and the agent plans it, writes it and runs its own hooks against your repository. What it does not have is any picture of how the code behaves once it ships.
The groundcover MCP server closes that gap. Point Kiro at your groundcover backend and the agent can read the same logs, traces, metrics and Kubernetes events your engineers read, scoped to a service and a time window, with no data leaving your cloud — and without changing how developers work.
What you'd use it for
- Debug the failure without leaving the editor.
Ask why checkout is erroring; the agent pulls the last hour of traces and error logs for that service and cites the spans it read. - Ground a spec in real traffic.
Before Kiro plans a refactor, it can check which endpoints actually carry load and what p99 looks like today, instead of guessing from the code. - Check a deploy before you call it done.
After a release, the agent compares error rate and latency against the previous window and reports the delta in the session.
How it works
The integration installs a local MCP server that talks to your groundcover backend over your service-account key. Kiro discovers its tools on session start; every call is read-only and scoped to the permissions on that key.
Before you start
- Kiro 0.9 or later, with MCP enabled in settings.
- A groundcover backend you can reach from your machine, and Node 20+.
- A read-only service account. Create it once and reuse the key; a personal token will break for teammates.
Observability for what comes next.
Start in minutes. No migrations. No data leaving your infrastructure. No surprises on the bill.
