Stop writing YAML by hand. Start describing what you need.
Quick Start • How It Works • Platforms • Documentation
You want a production PostgreSQL on AWS. Today, that means:
| Step | Traditional Approach | With InfraGenTool |
|---|---|---|
| 1️⃣ | 📖 Read Crossplane docs for hours | ✨ Describe what you need in plain English |
| 2️⃣ | 🔍 Hunt for the right provider CRDs and apiVersions | ✅ AI finds the correct schemas automatically |
| 3️⃣ | ✍️ Write 200+ lines of YAML by hand | ✅ Generate validated manifests instantly |
| 4️⃣ | 🐛 Debug cryptic validation errors | ✅ Zero hallucinated schemas — MCP-verified |
| 5️⃣ | 🔒 Hope you didn't forget security best practices | ✅ Security defaults built-in |
InfraGenTool eliminates all of this.
# Add the marketplace
/plugin marketplace add https://github.com/neelneelpurk/InfraGenTool.git
# Install the plugin
/plugin install infragentool@infragentool-marketplace
# Create a resource
/infragentool:new_composition postgres-db ./infra/databaseThe agent asks you a few questions, then generates:
./infra/database/
├── 📄 definition.yaml # XRD with full OpenAPI schema
├── 📄 composition.yaml # Composition with security defaults
├── 📄 claim.yaml # Ready-to-apply example claim
└── 📄 README.md # Usage documentation
📖 Full Getting Started Guide →
Three specialized AI agents work in sequence, each checking the previous one's work:
┌──────────────────────────────────────────────────────────────────────┐
│ │
│ YOU ──→ "I need a production Postgres with backups" │
│ │ │
│ ▼ │
│ ┌─────────────────────────┐ │
│ │ ☁️ Solutions Engineer │ ← Asks clarifying questions │
│ │ │ Writes the specification │
│ └────────────┬────────────┘ │
│ ▼ │
│ ┌─────────────────────────┐ │
│ │ 🏗️ Cloud Architect │ ← Reviews for security, │
│ │ │ cost, and reliability │
│ └────────────┬────────────┘ │
│ ▼ │
│ ┌─────────────────────────┐ │
│ │ ⚙️ Crossplane Engineer│ ← Generates YAML with │
│ │ │ MCP-verified schemas │
│ └─────────────────────────┘ │
│ │
└──────────────────────────────────────────────────────────────────────┘
Why three agents? The same reason engineering teams have architects review SWE proposals — separation of concerns catches mistakes that a single pass would miss.
# Hmm... is it rds.aws.crossplane.io or rds.aws.upbound.io?
# What's the apiVersion? v1beta1? v1beta2?
# Did I spell 'allocatedStorage' right?
# Is backupRetentionPeriod a string or integer?
# 😩 *3 hours later*# definition.yaml — Auto-generated XRD
apiVersion: apiextensions.crossplane.io/v1
kind: CompositeResourceDefinition
metadata:
name: xpostgresqls.infra.example.com
spec:
group: infra.example.com
names:
kind: XPostgreSQL
plural: xpostgresqls
claimNames:
kind: PostgreSQL
plural: postgresqls
versions:
- name: v1alpha1
served: true
referenceable: true
schema:
openAPIV3Schema:
type: object
properties:
spec:
type: object
properties:
parameters:
type: object
properties:
storageGB:
type: integer
instanceClass:
type: string
multiAZ:
type: boolean
backupRetentionDays:
type: integerEvery field name, apiVersion, and schema type is MCP-verified against the actual provider CRDs — never hallucinated.
| Command | Description |
|---|---|
/infragentool:setup |
Initialize project context (one-time) |
/infragentool:new_composition <name> <path> |
Create a new resource via multi-agent workflow |
/infragentool:update_composition <path> |
Update an existing resource (brownfield) |
/infragentool:implement_composition [track] |
Generate the YAML from an approved plan |
/infragentool:validate <path> |
Validate with crossplane render |
/infragentool:status |
Show all tracks and progress |
| Platform | Install Command | Context File |
|---|---|---|
| Claude Code | claude --plugin-dir ./InfraGenTool |
CLAUDE.md |
| Gemini CLI | gemini extensions install https://github.com/neelneelpurk/infragentool |
GEMINI.md |
| Spec Kit ⭐ | specify extension add infragentool |
SPECKIT.md |
All platforms share the same agents, knowledge base, and workflow — same quality output everywhere.
Using a different AI tool? If you're not on Claude Code or Gemini CLI, Spec Kit is the recommended approach. Spec Kit automatically supports multiple AI tools (Cursor, Copilot, Windsurf, and more), so you get full InfraGenTool support regardless of which assistant you use — no per-tool configuration needed.
infragentool/
├── .claude-plugin/plugin.json # Claude Code manifest
├── agents/ # AI agent definitions
│ ├── cloud_architect.md
│ ├── cloud_solutions_engineer.md
│ └── crossplane_engineer.md
├── commands/ # Agent command definitions
│ ├── setup.md # Claude commands (.md)
│ ├── new_composition.md
│ ├── update_composition.md
│ ├── implement_composition.md
│ ├── validate.md
│ ├── status.md
│ └── gemini/ # Gemini commands (.toml)
├── extension.yml # Spec Kit extension manifest
├── skills/infragentool/SKILL.md # Skill documentation
├── technical-docs/ # Cloud provider knowledge base
│ ├── cloud/ # AWS, Azure, GCP guides
│ └── crossplane/ # Crossplane patterns
├── CLAUDE.md # Claude Code context
├── GEMINI.md # Gemini CLI context
├── SPECKIT.md # Spec Kit context
└── GETTING_STARTED.md # Step-by-step tutorial
- Docker — Required for
crossplane rendervalidation - Crossplane CLI — Optional, for local validation
Contributions are welcome! Please read our Contributing Guide for details.
Licensed under the Apache License 2.0.