Build Infrastructure Context
for Your Coding Agent
Open-source CLI that syncs metadata from Grafana, Datadog, Kubernetes, and 20+ tools into structured markdown — so your AI agent stops hallucinating and starts debugging accurately.
Works with Claude Code, Cursor, Windsurf, and any agent that reads context files.
$ curl -fsSL https://raw.githubusercontent.com/DrDroidLab/context-builder/main/install.sh | bash
Also available via pipx install
With vs Without Infrastructure Context
Without droidctx
> Debug why payment latency spiked
Checking Kubernetes pods... (service is on ECS)
Writing PromQL from scratch... (dashboard exists)
Found redis-test-instance... (wrong instance)
Agent gave up after 15 tool calls.
With droidctx
> Debug why payment latency spiked
payment-service runs on ECS (from context)
Checking "Payment Service Overview" in Grafana
Redis prod metrics show p99 spike at 14:32
Upstream: api-gateway, Downstream: notification-svc
Root cause identified in 4 tool calls.
What Gets Generated
A structured directory of markdown files — click through to see what your agent gets.
Click files in the tree to preview their contents. This is generated from your live infrastructure.
Production Is Too Vast to Guess
A typical production setup has dozens of services spread across Kubernetes, ECS, and Lambda — with metrics in Grafana, alerts in Datadog, schemas in PostgreSQL, and deploys in ArgoCD. No agent can guess this topology.
Observability & infra tools
per team on average
Dashboards, alerts &
monitors to navigate
Of this is known to
your AI agent by default
Your agent is limited by what you know to tell it
Without context files, the agent can only navigate what you manually describe in your prompt. If you don't know that Redis metrics are in a specific Grafana dashboard, the agent won't either. It's restricted by your knowledge of where things live across a multi-tool system — which means it works for you, but can't be handed to a teammate who has different tribal knowledge.
Without a map, the agent wanders
When the agent doesn't know what exists — which services run where, which dashboards track what, how services connect — it guesses. It checks Kubernetes when the service is on ECS. It writes PromQL from scratch when a dashboard already has the panel. It finds a test Redis instance instead of production. Every wrong turn costs tool calls, tokens, and time.
How It Works
Four commands. That's it.
Initialize
Creates a credentials template for your infrastructure tools.
$ droidctx init Auto-Detect
Scans for kubectl, aws, gcloud, and az configs and auto-populates credentials.
$ droidctx detect Sync
Extracts metadata from all connected tools and generates structured markdown files with cross-references.
$ droidctx sync Point Your Agent
Reference the generated context in your CLAUDE.md or agent config. Your agent now has a complete map.
# CLAUDE.md
Reference ./droidctx-context/ for infra context 25+ Connectors
Syncs context from the tools your team already uses.
Grafana 
Datadog 
New Relic 
Kubernetes 
PostgreSQL 
MongoDB 
Elasticsearch 
AWS 
GitHub 
Sentry 
Jenkins 
ArgoCD 
CloudWatch 
Jira 
PostHog 
SigNoz 
GKE 
Google Cloud 
ClickHouse 
Prometheus 
Victoria Metrics 
Grafana Loki 
OpenSearch 
Coralogix 
Azure Give Your Agent Infrastructure Context
Install in 30 seconds. Generate context in one command. Debug accurately from day one.
$ curl -fsSL https://raw.githubusercontent.com/DrDroidLab/context-builder/main/install.sh | bash
$ droidctx init && droidctx detect && droidctx sync