ajeetraina777/sbx-grafana-kits repository overview

⁠Grafana kit for Docker Sandboxes

A standalone Docker Sandboxes⁠ kit (kind: mixin) that gives any sandbox agent access to Grafana⁠ — dashboards, datasources, and Prometheus/Loki queries — through the official Grafana MCP server⁠ (mcp-grafana), plus the grafana-client Python library. This image ships in three target flavors, one per tag.

Source and full docs: https://github.com/ajeetraina/sbx-kits-grafana⁠

⁠Image tags

Local is the default because it needs no Grafana account and no token: bring up Grafana on the host and the sandbox reaches it over host.docker.internal. Cloud and self-hosted users authenticate with a Grafana service-account token, stored once with sbx so it never enters the sandbox.

⁠Quick start

Local default. Bring up Grafana on the host, then launch:

docker compose -f compose/docker-compose.yml up -d   # from the repo
sbx run --kit docker.io/ajeetraina777/sbx-grafana-kits:latest claude

Grafana Cloud. Store a service-account token once (never on the command line), set your stack URL (see the repo's providers/cloud.md), then run:

echo "$GRAFANA_TOKEN" | sbx secret set -g grafana
sbx run --kit docker.io/ajeetraina777/sbx-grafana-kits:cloud claude

Self-hosted:

echo "$GRAFANA_TOKEN" | sbx secret set -g grafana
sbx run --kit ./kits/oss claude   # edit kits/oss/spec.yaml with your host first

The cloud/oss tags hold no token. The sbx proxy injects it from the stored secret, so the token never enters the sandbox. sbx run has no -e flag by design.

⁠How it works

Each kit installs the pinned mcp-grafana v1.0.0 binary and grafana-client, sets GRAFANA_URL, writes a portable MCP definition to ~/.grafana/mcp.json, and (best-effort) registers the server with the Claude agent. The cloud/oss tags additionally route requests to your Grafana host through the sbx proxy with an Authorization: Bearer header sourced from the stored grafana secret. It also ships a runbook (~/runbooks/grafana_report.py) that prints the instance's health, datasources, and dashboards.

Per-target setup notes, validation steps, and the raw spec.yaml for each kit live on GitHub: https://github.com/ajeetraina/sbx-kits-grafana/tree/main/providers⁠